
Royal Courts of Justice
Strand, London, WC2A 2LL
Before :
MR JUSTICE JOHNSON
Between :
(1) Wikimedia Foundation | ||
(a charitable foundation registered in the United States of America) | ||
(2) BLN | Claimants | |
- and - | ||
Secretary of State for Science, Innovation and Technology | Defendant | |
Rupert Paines and Raphael Hogarth (instructed by Pinsent Masons LLP) for the Claimants
Cecilia Ivimy KC, Ruth Kennedy and Joseph Lavery (instructed by the Government Legal Department) for the Defendant
Hearing dates: 22 – 23 July 2025
Approved Judgment
This judgment was handed down by release to The National Archives on 11 August 2025
Mr Justice Johnson:
The first claimant is a charitable foundation which hosts the online encyclopaedia, Wikipedia. The second claimant is a user and editor of Wikipedia. The claimants challenge the Secretary of State’s decision to make regulation 3 of the Online Safety Act 2023 (Category 1, Category 2A and Category 2B Threshold Conditions) Regulations 2025. That regulation prescribes the criteria for “Category 1” online services, which are made subject to many statutory duties.
The claimants say that the criteria are logically flawed. They were intended to capture large profitable social media companies where anonymous content can “go viral”. However, they were (say the claimants) drawn too broadly with the result that Wikipedia is likely to qualify as a Category 1 service even though that was never the policy intention. If Wikipedia is a Category 1 service, that will fundamentally change the way it operates. Either the number of people in the United Kingdom who access Wikipedia will have to be reduced by around three quarters, or important functionality corresponding to Category 1 threshold criteria must be disabled, or else the first claimant will have to comply with duties that are not reasonably manageable and which are incompatible with the way in which Wikipedia operates.
Allowing for the way in which the arguments were ultimately advanced (and slightly re-numbering and re-characterising the grounds accordingly), the claimants say that the decision to make the regulation was flawed on the grounds that:
the Secretary of State failed to comply with a duty imposed by paragraph 1(5) of schedule 11 to the Online Safety Act 2023 to take into account the likely impact of the number of users of the user-to-user part of a service, and its functionalities, on the ease, speed and breadth of the dissemination of user-generated content.
the decision was irrational because it was based on flawed reasoning.
the decision is incompatible with articles 8, 10 and 11 of the European Convention on Human Rights.
the decision is incompatible with article 14 of the Convention (or is otherwise irrational) because it fails to distinguish between different types of online provider.
There is a degree of urgency in these proceedings because they are, in effect, holding up part of a complex regulatory process that needs to be worked through before important parts of the Act take effect. For that reason, the claim was listed for hearing on an expedited basis for the court to consider both the question of whether permission should be granted to bring a claim for judicial review and, if so, whether the decision to make regulation 3 was flawed on one or more of the grounds asserted by the claimants.
BLN has filed a witness statement, as has Philippe Bradley-Schmieg who is the first claimant’s in-house lead counsel. The defendant has filed a statement from Talitha Rowland, Director, Security and Online Harms at the Department for Science, Innovation and Technology. Ms Rowland has been in her current post throughout the period when consideration was given to the making of the 2025 regulations. She provided senior civil servant sign off on all relevant Ministerial Submissions, and she engaged regularly with the policy team and with ministers and their officials throughout the decision-making process.
Wikipedia
Mr Bradley-Schmieg explains how Wikipedia works.
Wikipedia is a free, public interest, non-profit, collaboratively edited encyclopaedia that is available to anyone with access to the internet anywhere in the world, save for a small number of countries or institutions that block access. There are over 300 different language versions. It is said to be the largest and most-read reference work in history. Over 5 billion pages on the English language version of Wikipedia are viewed by users each month. That includes 789 million pages that are viewed by users in the United Kingdom. It is accessed by around 26 million people in the United Kingdom each month.
Wikipedia was founded in January 2001 by Jimmy Wales and Larry Sanger. It is hosted by the first claimant, which provides the software for it to function as well as the necessary technical, administrative and legal infrastructure. It operates in a highly transparent and democratic manner. The underlying computer code is open-source and can be freely inspected on the internet. Its policies and guidelines are developed by its users and are, again, published. Wikipedia runs on software that enables visitors to its website to create and edit content. That in turn enables content to be modified, curated or deleted by users. The result is that with very limited exceptions (for example, removal of content required by a court order, and automatic removal of child sexual exploitation and abuse material) the content is entirely self-regulated by users.
Any user can add an article or amend an existing article. As soon as they do so, the new or amended article is (usually) instantly available to all other users, who can then, in turn, edit that new article. This also means that any one article may be the product of the work of many different users. For example, an article relating to Queen Elizabeth II has been edited over 18,000 times, and the first sentence alone is the product of edits made by 11 separate authors. Mr Bradley-Schmieg says this means that the user verification duties in the Act (under which users can choose only to encounter content from other users whose identity has been verified) are incompatible with the nature of Wikipedia as a collaboratively edited encyclopaedia. Content that has been generated by users whose identity has been verified cannot sensibly be isolated from content generated by anonymous users without rendering the service unusable: articles would, generally, not make sense.
The community of users sets the principles and policies that are applied to content on Wikipedia. They include neutrality of viewpoint and an obligation to provide references to reliable and publicly verifiable sources. If a user does not comply with Wikipedia’s principles and policies, then their edits are likely to be deleted by other users. This process of user-moderation works. Content that does not comply with Wikipedia’s principles and policies is rapidly removed. An independent study has shown that the median time for such content to be taken down is just 61 seconds. Harmful content is not typically encountered. In 2024, the first claimant received 664 requests to take down material (across all of its projects, not just Wikipedia), and it was required to do so in only 4 cases. By contrast, TikTok received 375,899 reports of illegal content in the same period and removed 63,723 items.
Mr Bradley-Schmieg explains that Wikipedia is different from large social media companies in respect of its legal governance (as a non-profit charity), its (lack of a) business model, the fact that its users typically only encounter content that they seek out, the first claimant’s very limited role in moderation, and (partly as a result of those features) its low risk profile. Ofcom’s research indicates that of users who experienced online harm, 56% was due to social media, 9% was due to sites that host videos posted by other users, and 8% was due to webmail. Only 2% was due to a general “other” category (which includes Wikipedia).
Mr Bradley-Schmieg sets out the practical implications if Wikipedia is subject to the Category 1 conditions. Because of the way in which Wikipedia operates, it has little spare resource and it is already over-stretched because of existing regulatory burdens. The user verification duties alone would require Wikipedia to build new systems, data handling processes, functionality and tools. It would face huge challenges to meet the large technological and staffing needs that would have to be deployed to meet these statutory requirements. Other Category 1 duties would also have a significant impact. Mr Bradley-Schmieg “do[es] not see how the… duties could be reasonably manageable for [Wikipedia].” In practice, it is likely that Wikipedia would have to consider taking steps to ensure that it does not fall within the scope of a Category 1 service. That might involve preventing access once a monthly limit has been reached (in effect restricting access to just one quarter of those who currently use Wikipedia in the United Kingdom) or removing functionalities and characteristics that are important to the way that Wikipedia functions.
BLN is a user of Wikipedia. Initially, they used it to browse articles of interest. Eventually, they started editing articles by correcting the occasional typing error. With time and experience they began to make more substantial edits. BLN has now become a regular editor, to the point that they have made almost 140,000 edits, from correcting typing errors to adding references, contributing additional information and updating content. They also help other users. In 2007, BLN was elected by other Wikipedia editors to the role of administrator. There are currently 847 administrators. They are able, in accordance with detailed policies, to block users and IP addresses, delete pages that do not comply with Wikipedia’s policies, and protect pages from editing where there is a risk that edits will not comply with those policies. In 2017, BLN was selected by other users to act as a “checkuser”. There are currently 54 checkusers. They can access (subject to a non-disclosure agreement) technical data from Wikipedia’s servers that relate to individual users or IP addresses. The checkusers can establish if two or more accounts are being operated by a single person or group for malicious or forbidden reasons and can protect Wikipedia against disruptive or abusive behaviour.
BLN gives a practical example of the way in which users self-regulate Wikipedia’s content. Following the attack perpetrated by (as it turned out) Axel Radukubana in Southport on 29 July 2024, there was much speculation on social media as to the identity of the perpetrator. That is said to have contributed to subsequent public disorder. Some Wikipedia users edited a page about the attack to include what was (often wrongly) said to be information about the attack, including the perpetrator’s name, but without any cited authoritative source. When this happened, other Wikipedia users rapidly removed material that was not authoritatively sourced. After some instances of this, BLN temporarily applied “page protection” to the article to prevent it from being edited by inexperienced users. This was done on 30 July 2024. The page was then closely monitored by several editors to ensure the reliability of any changes. Also, a warning was applied to explain that it contained breaking news and should be approached with caution.
BLN says:
“Wikipedia [has always had] a bold and valuable ambition, to let members of the public come together to create a high-quality free global repository of knowledge. I firmly believe that more access to information allows individuals and societies to make better decisions. It also empowers individuals to pursue their goals and interests, whether that is through research in an educational or career setting, by sharing random facts with friends, or plain curiosity. Wikipedia also has an essential feature, which is that it is published under a free copyright licence, meaning that third parties are lawfully permitted to reuse both Wikipedia content and software, for free, under permissive intellectual property licenses. This prevents contributions and knowledge being restricted or monopolised, and allows maximum opportunity for content to develop, and then to be shared with others. Wikipedia is not only a free resource for facts. Through its policies of open participation and presenting multiple points of view, it is a gateway to learning and being critical about knowledge and different ways of thinking - both as a reader or as an active contributor.”
BLN gives detailed and compelling reasons for wishing to remain anonymous as a Wikipedia user. They provide extensive evidence of threats that are made to Wikipedia administrators, including to BLN personally, on websites that target Wikipedia administrators. They also explain the significant and detrimental effect on Wikipedia if Category 1 conditions applied. If a quota system were introduced to avoid the conditions then that would limit their (or others) ability to access an excellent source of information, reduce the quality of the information they could access, limit their ability to help with editing, moderation and administration, and effectively prevent them from participating in the life of the Wikipedia community to which they have invested so much time and energy over many years. If editors were required to verify their identity, then they would be faced with an invidious choice between providing their identity to Wikipedia (and thereby compromising their privacy and, potentially, safety in the event of a hack or data leak) or ceasing some of the important activities they perform to ensure that Wikipedia’s content complies with its policies.
None of the evidence of Mr Bradley-Schmieg or BLN was challenged. Nor was any evidence advanced that Wikipedia poses any form of threat to freedom of expression or informed public discourse or, specifically, the public interests that are protected by article 10(2) of the Convention (which can, in principle, justify a proportionate interference with free expression rights). The unchallenged evidence, therefore, is that Wikipedia is a tool that provides significant value for freedom of speech and expression, particularly the right “to receive and impart information and ideas without interference by public authority and regardless of frontiers” (to use the language of article 10(1) of the Convention) and that it does so without giving rise to any substantial threat to the public interests that are protected by article 10(2) of the Convention.
It follows that any decision by a public authority (including a decision to make (or not to amend) secondary legislation) that has a significant impact on Wikipedia’s ability to operate would, in the absence of justification, likely be unlawful as being contrary to section 6 of the Human Rights Act 1998 read with article 10 of the Convention.
The 2023 Act provides for a new regulatory framework to make the use of certain internet services safer for individuals: section 1(1). It does that by imposing duties on the providers of those services and by conferring functions and powers on Ofcom: section 1(2) and Part 7. Those powers include powers to gather information (including by way of powers of entry, inspection and audit): Chapter 4 of Part 7. The duties imposed by the Act seek to secure that such services are “safe by design” and are designed and operated in a way that protects users’ rights to freedom of expression: section 1(3)(a), (b)(ii).
One type of service that is regulated by the Act is a “user-to-user service.” That is a service by means of which content generated by one user may be encountered by another user: section 3(1). A user-to-user service is a regulated service, and a “Part 3 service”, if (subject to exceptions which do not matter for the purposes of this claim) it has a significant number of users in the United Kingdom: section 4(2), (3), (4), (5)(a) read with section 227(1)(a). With limited prescribed exceptions (such as emails and SMS messages), all user-generated content on a regulated user-to-user service is itself regulated: section 55(2).
The Act imposes duties on all regulated user-to-user services. These include duties to assess, mitigate and manage risks posed by certain types of illegal content, enabling users to report illegal content, the provision of a complaints procedure and duties to have regard to freedom of expression and privacy when implementing safety measures and policies: sections 9 – 10 and 20 – 23, read with section 59.
The Act applies additional duties to different categories of service. The most intensive duties and oversight apply to “Category 1” services. When the Bill was first introduced, it had separate provisions that were designed to protect adults from “legal but harmful” content. Those provisions were removed but were replaced by the duties that apply to “Category 1” providers. They include:
Duties to give users a choice about whether to verify their identity and the type of content they see, including whether they see content from users who have not verified their identity: sections 14, 15 and 64.
Duties to protect free speech: sections 17 – 19.
Duties to protect users from fraudulent advertising: section 38.
Duties to ensure compliance with the service’s terms (including that the service provider takes down user-generated content if but only if that content does not comply with the terms of service): sections 71 and 72.
A duty to provide and publish an annual “transparency report” to Ofcom containing such information as Ofcom requires: section 77.
Miscellaneous additional duties, including to summarise the service’s most recent illegal content risk assessment and children’s risk assessment and to make provision for a complaints procedure: sections 10(9), 12(14) and 21.
Many of the duties are couched in terms of proportionality. For example, the duty under section 17(2) to protect content of democratic importance is expressed as a “duty to operate a service using proportionate systems and processes designed to ensure that the importance of the free expression of content of democratic importance is taken into account when taking decisions about (a) how to treat such content… and (b) whether to take action against a user generating… such content.”
However, not all duties are couched in terms of proportionality. One which is not, and about which the first claimant is particularly concerned, is a duty to enable users to filter out content from non-verified users – see section 15(1), (9) and (10):
“15 User empowerment duties
(1) This section sets out the duties to empower adult users which apply in relation to Category 1 services.
…
(9) A duty to include in a service features which adult users may use or apply if they wish to filter out non-verified users.
(10) The features referred to in subsection (9) are those which, if used or applied by a user, result in the use by the service of systems or processes designed to effectively—
(a) prevent non-verified users from interacting with content which that user generates, uploads or shares on the service, and
(b) reduce the likelihood of that user encountering content which non-verified users generate, upload or share on the service.”
The Secretary of State has the power to exempt content or services from the Act, including a power to provide for a description of user-to-user service to be exempt if the Secretary of State considers that the risk of harm is low: section 220(4).
Ofcom is required to prepare Codes of Practice which set out the measures that providers may take to comply with certain Category 1 duties: section 41. In doing so, it must consult persons with expertise in the right to freedom of expression: section 41(6)(f)(i). Category 1 duties do not apply until the applicable Code of Practice is in force: section 51. The measures prescribed by the Codes of Practice must be proportionate and technically feasible: paragraph 1 of schedule 4. They must incorporate safeguards for freedom of expression and privacy: paragraph 10 of schedule 4. If a service provider complies with the applicable Code of Practice they are deemed to have complied with the statutory duty: section 49(1). Ofcom must publish guidance in respect of those Category 1 duties which are not subject to a Code of Practice: section 52.
Ms Rowland says that Category 1 duties are aimed at securing freedom of expression by protecting certain types of content that are important to public discourse and securing that services are transparent and accountable. They are intended to apply to a range of different kinds of service, but the impact of the duties will vary. Thus, if a service does not host advertising it may not need to take any measures to protect users from fraudulent advertising.
The Act prescribes an elaborate step-by-step process for the making of regulations that specify the Category 1 threshold conditions and, thereafter, their implementation: sections 94 – 95 and schedule 11.
First, within six months of the Act coming into force, Ofcom was required to carry out research into (a) how easily, quickly and widely regulated user-generated content is disseminated by means of regulated user-to-user services, (b) the number of users and functionalities of the user-to-user part of such services, and (c) such other characteristics of that part of such services or factors relating to that part of such services as Ofcom considered to be relevant to specifying the Category 1 threshold conditions: schedule 11, paragraph 2(2). The concept of easy, quick and wide dissemination of regulated user-generated content by means of regulated user-to-user services may be seen as a codification of the well-known concept of user-generated content (particularly on social media) “going viral”. For shorthand, I will refer to this as “viral dissemination”. A “functionality” is a feature that enables interactions of any description between users, including, for example, features that enable users to create a profile or search for user-generated content or share content with others: section 233.
Second, Ofcom must then provide the Secretary of State with advice based on that research as to the provision which Ofcom consider it is appropriate for regulations prescribing the Category 1 conditions to make: schedule 11, paragraph 2(5).
Third, as soon as reasonably practicable after the provision of that advice, Ofcom must publish the advice: schedule 11, paragraph 2(7).
Fourth, as soon as reasonably practicable after the provision of that advice, the Secretary of State is required to lay a draft of the statutory instrument containing the proposed regulations before each House of Parliament: schedule 11, paragraph 2(7), and section 225(8).
Fifth, the Secretary of State is then required to make “the first regulations” under paragraph 1(1) of schedule 11: schedule 11, paragraph 2(7).
Sixth, Ofcom is then required to establish a register of Category 1 services as soon as is reasonably practicable: section 95(1).
Seventh, Ofcom is then required to publish Codes of Practice and guidance about the duties owed by providers under the Act, including duties owed by Category 1 providers: Chapter 6 of Part 3.
Eighth, from the point that the first regulations are made, Ofcom may conduct further research, whether on its own initiative or at the request of the Secretary of State, to assess whether to amend the regulations: schedule 11, paragraphs 3(1), (4).
The regulations that specify the conditions for qualifying as a Category 1 service must relate to (a) the number of users of the user-to-user part of the service, (b) functionalities of that part of the service, and (c) any other characteristics of that part of the service or factors relating to that part of the service that the Secretary of State considers relevant: schedule 11, paragraph 1(1). The regulations must specify the ways in which the Category 1 threshold conditions may be met, and they must provide that at least one specified condition about number of users or functionality must be met: schedule 11, paragraph 1(4).
Schedule 11, paragraph 1(5) is the foundation for the claimant’s first ground of claim. It states:
“In making [the] regulations… the Secretary of State must take into account the likely impact of the number of users of the user-to-user part of the service, and its functionalities, on how easily, quickly and widely regulated user-generated content is disseminated by means of the service.”
Ofcom’s research and advice
The Act came into force on 26 October 2023. Ofcom was then required to conduct its research within 6 months. That was the first stage in the statutory process for making the regulations.
Ofcom’s research was informed by four themes: consistency (applying the same data sources and research for category 1 conditions as for other category conditions), objectivity (the underlying work was done in a “service-agnostic” way, in part by ensuring that data could not be linked to an identified service), scope (the aim was to use comprehensive and reliable data sources, recognising the lack of robust data on matters such as user numbers), and transparency (by publishing the data and additional research sources).
In conducting its research into viral dissemination, Ofcom paid particular attention to different functionalities (including different forms of livestreaming, different types of forwarding or re-sharing user-generated content, and in-livestream chat) which it considered facilitated users to disseminate material near-instantaneously, with the potential to reach multiple users at once. For these purposes, it treated “content recommender systems” as a functionality (albeit it says that they do not strictly meet the statutory definition of a functionality). It considered the operation of these functionalities across many different types of service, including social-media, information-sharing, pornography, marketplaces and video-sharing.
It summarised its research thus:
“Research: Our research findings indicate that the features of a service most relevant to content being disseminated easily, quickly and widely are:
• content recommender systems, because they are typically relied upon by services to amplify content to a wide set of users; and
• the ability for users to forward or re-share content, because this facilitates users sharing content instantaneously with others.
In our view, these two features each operate to increase dissemination of content easily, quickly and widely. Additionally, the effects of these features are likely to be increased further as the user base increases and when these features operate in combination. In essence, the higher the user base, the more content that is likely to be shared.”
The second stage was for Ofcom to provide advice to the Secretary of State. It did that on 29 February 2024. It provided the following summary of its advice:
“Advice: For these reasons, our advice is that category 1 thresholds should target services that fulfil either of the two following sets of conditions:
Condition 1:
• the use of a content recommender system on its service; and
• have more than 34 million UK users on the user-to-user part of the service, representing c.50% of the UK population.
Condition 2:
• have a functionality that offers users the ability to forward or re-share user-generated content with other users of the service; and
• the use of a content recommender system on its service; and
• have more than 7 million UK users on the user-to-user part of the service, representing c.10% of the UK population.”
In the body of its advice, it explained its rationale:
“3.18 We analysed each of the six functionalities and content recommender systems and have concluded that two, in particular, stand out from our research as playing a particularly significant role in the dissemination of regulated user-generated content. These are:
a) the use of a content recommender system; and
b) the ability to forward or re-share user-generated content with other users of the service.
3.19 Content recommender systems amplify the breadth, scale and speed of content dissemination on a service by proactively disseminating content to new users or groups of users. Content recommender systems therefore play a fundamental role in enabling content to be disseminated easily, quickly and widely. Forwarding or re-sharing existing content with other users of a service is another key component of content dissemination, as it allows for the movement of existing content to new users or groups of users, by specifically enabling users themselves to affect content dissemination. These two features allow both the service and the user to affect the dissemination of content.
…
3.23 …we judge that where services have a very large number of users, a content recommender system alone is sufficient for content to be disseminated easily, quickly and widely. This is because a service’s content recommendation system results, in and of itself, in the dissemination of content to a very large audience without the need for users themselves to further share content.
3.24 Where services have a lower but still considerable number of users, a content recommender system alone may not be sufficient to disseminate content quickly, easily and widely. We judge that the ability for users to forward or re-share existing content on the service operates in conjunction with a content recommender to increase the likelihood of quick, easy and wide content dissemination: this can occur both through active content dissemination, driven by users' own engagement with content, as well as via content dissemination driven by the service's own systems.
3.25 Based on our analysis above, we therefore consider it appropriate to recommend two sets of thresholds for category 1.
3.26 We are required to recommend a user number threshold to the Secretary of State. In doing so we have exercised our regulatory judgement having regard to our general duties under the Communications Act 2003 and the function we are carrying out.
3.27 Set one: Considering the quick, easy and wide dissemination of content, we judge that services with content recommender systems and very large user bases are relevant to category 1. Based on our analysis, we propose that a user number threshold of 34 million UK users of the user-to-user part of the service is appropriate for such services. This represents approximately 50% of the total UK population.
3.28 Set two: Considering the quick, easy and wide dissemination of content, we judge that services with content recommender systems, the ability for users to forward or re-share existing content on the service and large user bases are relevant to category 1. Based on our analysis, we propose that a user number threshold of 7 million UK users of the user-to-user part of the service is appropriate for such services. This represents approximately 10% of the total UK population.
3.29 Our preliminary indicative analysis suggests that approximately 12-16 services may meet one or both of these user number thresholds, when factoring in the impact of the functionality requirements described above. This estimated number of services in our view indicates that our recommended user number thresholds are likely to strike the right balance in terms of targeting those services where content is likely to be disseminated easily, quickly and widely, while ensuring that the duties apply to a sufficiently targeted number of services.”
The Secretary of State’s decision to make regulation 3
The third stage was that Ofcom was required to publish its advice. This was to happen as soon as reasonably practicable after Ofcom advised. Ofcom published its advice on 25 March 2024 (the advice having been provided to the Secretary of State on 29 February 2024).
The fourth stage was for the Secretary of State to make draft regulations prescribing the Category 1 conditions. Again, this was to happen as soon as reasonably practicable after Ofcom advised.
Between March 2024 and February 2025 officials prepared 14 Ministerial Submissions in relation to the making of the regulations. During this period there was a change of Government. The General Election was called on 22 May 2024 and the new Government was formed on 7 July 2024. The new Government was not shown earlier Submissions, but the work that had been done by officials in the period between March and July 2024 informed the advice that was given to the new Government.
On 7 March 2024, the then Secretary of State was asked to decide whether to accept Ofcom’s recommended thresholds, so that work could commence on drafting the relevant secondary legislation. Officials recommended that she do so. Initially, the Secretary of State was not content with the advice. She was particularly concerned that “the likes of Amazon was likely to be captured by Category 1… but not Pornhub… [and that] small but risky services are not covered.”
On 18 March 2024, the Secretary of State was provided with a Submission which made it clear that Category 1 duties were not primarily aimed at pornographic content or the protection of children (which were dealt with by other parts of the Act). Rather, the aim of Category 1 was to capture services that have a significant influence over public discourse. The submission offered, as a possible option, requesting information from Ofcom as to how content recommender systems function on different types of service. It was envisaged that this might indicate that that there were differing levels of impact on viral dissemination, and that could be used to carve out certain content recommender systems from the ambit of the regulation.
Officials then spoke to Ofcom policy directors who advised that there was very little information available on how content recommender systems work across different types of service, and that it would be extremely difficult to make robust regulations that differentiated between different types of service. On 21 March 2024, a further submission was put before the Secretary of State in which it was said that Ofcom did not have additional information about how content recommenders work on different services, so that would require additional work which would impact on the timeline for laying regulations. It was therefore suggested that the Secretary of State should not ask for more information about this.
On 9 April 2024, the Secretary of State received a letter from “the Mid Size Platform Group”. This is a group that comprised seven (subsequently eight) online services, including the first claimant. The letter captured what, in these proceedings, the first claimant say is a key problem with regulation 3(2):
“Definition of recommender systems: …the proposed definition is too broad and will encompass essentially any platform in the sector which organises content in any way. Content recommender systems are diverse in their design and application. While we recognise some content recommender systems are designed to encourage users to chase sitewide virality, which may lead to negative outcomes, others are designed for safety reasons, for marketplace efficiency, or are simply an ancillary part of the service. For example, they can be used to ensure that content deemed safe, age appropriate, or produced by a user with a positive safety track record is more visible to users. In addition, it would catch services that allow users to access recommended content but do not force them to use an algorithmic feed if they prefer not to. We therefore recommend that the definition of recommender system is made more detailed and nuanced, rather than simply the presence of the functionality, to avoid catching a high number of less risky platforms in Category 1. For example, systems with a safety component, systems merely reflecting past purchase history, as well as systems that are a minor and ancillary component of the service, should be exempted.”
On 22 May 2024 (the day on which the general election was called), departmental officials held a workshop with Ofcom. By this point, the Secretary of State had indicated that she agreed with officials that she should follow Ofcom’s advice. A paper circulated in advance of the workshop said that the proposed definition of a content recommender system had been criticised for being too broad. It raised the following, among other, questions:
“• Are there any recommender systems that Ofcom would discount in line with the definition?
• Wikimedia raised concerns that the broadness of the definition would capture systems such as this [the Wikipedia “New Pages Feed”] - does Ofcom agree with this?”
The record of the discussion that took place at the workshop does not directly address this issue. However, during the election period, officials considered the concern that the definition of a content recommender system was too broad. Options were discussed which would exclude Wikipedia. They included excluding charitable services, excluding services without paid-for advertising, narrowing the definition of a content recommender system to only include systems that are “integral” to the operation of the service, amending the definition of “users” to include only those that interact with the content recommender system, and commissioning further research from Ofcom. Officials concluded that introducing a change without commissioning further research from Ofcom could lead to unintended outcomes. In particular, the opacity of content recommender systems meant that narrowing the definition could lead to the exclusion of services that ought to be included, and commissioning further research would lead to significant delay. Also, the pace of technological developments was such that any further research would require continual reassessment.
On 5 July 2024, a new Government was formed. Ms Rowland says that the work that had been conducted up to that point fed into the material that was provided to the new Secretary of State. She says he was extremely engaged with the setting of the Category 1 thresholds and that she had a number of discussions with him and/or his Special Advisors about the issues.
On 18 July 2024, the new Secretary of State was first provided with a formal Ministerial Submission in relation to the regulations. He was told that Ofcom’s recommendations were “broadly in line with… expectations… in terms of the number and type of services in scope” and that “[w]hilst there are some outliers [and elsewhere in the document Wikipedia was identified as an outlier]… the services likely to be in scope are largely justifiable…” He was informed that some companies, including Wikipedia, had criticised the thresholds and believed that the recommendations were too broad, and that many parliamentarians were particularly concerned about Wikipedia being in scope. It was then said, “Wikipedia has content of democratic important and journalistic content…” Officials recommended following Ofcom’s advice. The Secretary of State was given a list of services that were likely to be captured by the proposed category 1 thresholds, and also a list of services “which could possibly be captured”. Wikipedia was in the latter list. Officials advised that they had not expected Wikipedia to be in scope, but it had raised the possibility that it could be due to the proposed definition of a content recommender system.
The Secretary of State was informed that he was required to take into account the likely impact of the number of users of the service, and its functionalities, on viral dissemination, as well as any other characteristics or factors that the Secretary of State considered relevant. He was told that Ofcom had concluded that content recommender systems played a particularly significant role in the dissemination of content because they are typically relied on by services to amplify content to a wide set of users, and that:
“Ofcom judged that where services have a very large number of users, a content recommender system alone is sufficient for content to be disseminated easily, quickly and widely. This is because a service’s content recommendation system results, in and of itself, in the dissemination of content to a very large audience without the need for users themselves to further share content.”
Attention was drawn to limitations on the evidence and research that informed Ofcom’s advice. Attention was also drawn to the concerns expressed by the Mid-Size Platform Group, but it was said that it would be difficult to tailor the thresholds to meet its concerns. It was pointed out that Ofcom had not been able to provide more granular detail about how different content recommender systems worked in the context of different services, that they were often opaque and that they changed frequently, that Ofcom wanted to do further work to understand that better, but that it would need to use its information gathering powers to do that.
A Departmental Minister, Baroness Jones, asked for a meeting to discuss the Submission with officials. That took place on 25 July 2024. Baroness Jones expressed concern that certain “outliers” might fall within the scope of Category 1 services, particularly Google Maps, Google Photos, Amazon and eBay (but was apparently less concerned about the potential inclusion of Wikipedia). Following that meeting, officials considered options for excluding certain “outlier” services, including Wikipedia.
On 31 July 2024, a further Ministerial Submission recommended following Ofcom’s advice. Options were put forward which could potentially have the effect of excluding certain outlier services (including online encyclopaedias). Officials advised that these options were not recommended because of the risk (in the absence of further research) of inadvertently creating loopholes. It was recognised that further research could be conducted into viral dissemination and to understand more about content recommender systems. However, Ofcom had indicated that they did not have a statutory power to conduct such further research (but that they would be able to do so once the first regulations were in force). It was also said that there were policy reasons why online encyclopaedias should be within the scope of Category 1, and that Wikipedia held content of democratic importance and journalistic content. The options that were presented included amending the definition of a content recommender system, but it was said that it was difficult to do this in a way that was objective, and evidence based, and which did not create any risk of loopholes. In providing this advice, officials drew on the work that had been done, including engagement with Ofcom, over the previous months.
On the same day, the Minister responded, “I reluctantly agree with the proposals in the submission as I recognise that our hands are largely tied by the constraints of the Act.” The Minister also directed officials to undertake stakeholder liaison.
A meeting then took place on 12 August 2024, attended by the Secretary of State, the Minister and officials. The Secretary of State considered the question of online encyclopaedias but was content not to seek further advice. He indicated that he was broadly content to accept Ofcom’s recommendations, subject to an issue concerning suicide forums. That issue was addressed in a further Submission. On 17 September 2024, the Secretary of State agreed with the recommendation to approve Ofcom’s proposed conditions.
Ms Rowland correctly observes that Ofcom did not provide advice about the specific impact of each functionality or characteristic of each service on that service’s dissemination of content. Its research was undertaken at a higher level of generality and considered how different functionalities and characteristics worked across a range of different types of service, and their impact – across that range – on viral dissemination. Its advice was provided on that same basis, and that was how the Secretary of State understood it. He was not, therefore, under a misapprehension that content recommender systems worked in the same way across different types of service, or that they affected viral dissemination to the same degree. He was advised that they operated in an opaque manner that was subject to rapid change, but there was a limit to the advice that could be provided without Ofcom exercising its information gathering powers.
A draft of the regulations was laid before Parliament on 16 December 2024. It was approved by a resolution of each House of Parliament, as required by section 225(8) of the Act. The Secretary of State then made the regulations on 26 February 2025. They came into force the following day.
Regulation 3 states:
“3 Category 1 threshold conditions
(1) The Category 1 threshold conditions are met by a regulated user-to-user service where, in respect of the user-to-user part of that service, it—
(a) (i) has an average number of monthly active United Kingdom users that exceeds 34 million, and
(ii) uses a content recommender system, or
(b) (i) has an average number of monthly active United Kingdom users that exceeds 7 million,
(ii) uses a content recommender system, and
(iii) provides a functionality for users to forward or share regulated user-generated content on the service with other users of that service.
(2) In paragraph (1), a “content recommender system” means a system, used by the provider of a regulated user-to-user service in respect of the user-to-user part of that service, that uses algorithms which by means of machine learning or other techniques determines, or otherwise affects, the way in which regulated user-generated content of a user, whether alone or with other content, may be encountered by other users of the service.”
The next steps
The next step is for Ofcom to establish a register of Category 1 services. To do this, it must identify which services fall within the scope of regulation 3. In doing so, it must (so far as possible) read and give effect to regulation 3 in a way which is compatible with Convention rights: section 3(1) of the Human Rights Act 1998.
In order to determine which services meet the Category 1 criteria, Ofcom has issued information notices to providers of services that it considers might do so. One such notice was issued to the first claimant on 27 March 2025 (a draft having been sent in January 2025). It asks the first claimant to provide the mean number of monthly active United Kingdom users for Wikipedia, how it determines whether a person is an active United Kingdom user, and whether Wikipedia uses a content recommender system and whether it provides a functionality for users to forward or share regulated user-generated content with other users. The first claimant responded on 16 May 2025. It contended that Wikipedia did not fall to be categorised as a Category 1 service. It pointed out that insofar as Wikipedia has a content recommender system and forward/share functionality, this is not typically used to serve content to large numbers of users in any significant way, it is not integral to the user “journey”, and it is primarily used for ancillary purposes such as content moderation. Ofcom has not yet determined whether Wikipedia is a Category 1 service.
Once Ofcom has published the register of Category 1 services, it will issue draft transparency notices “within a few weeks” and will then issue final transparency notices “soon thereafter”. It will then publish draft Codes of Practice. Ms Rowland, in her witness statement, said that would be “no later than 2026”. I was told that the Codes of Practice would not be in force before 2027.
The Secretary of State may, at any stage, amend the regulations (subject to certain procedural requirements, including taking further advice from Ofcom). In particular, he may do so once Ofcom has established its register of Category 1 services. Thus if, at that point, services are included that the Secretary of State considers should not be included, it is open to the Secretary of State to amend the regulations. Arguably, the Secretary of State would be under a statutory duty to do so if that were necessary to avoid a breach of Convention rights. Ms Rowland, in her witness statement, says that when the Secretary of State made regulation 3 he took account of his power to amend them in the light of subsequent developments when the impact of imposing Category 1 duties would be better understood. He also took account of the relatively long lead-in time to the implementation of Category 1 duties.
Application of the Category 1 threshold conditions to Wikipedia
Viral dissemination is not a concept that obviously applies to Wikipedia. The typical Wikipedia user only encounters content that they seek out. Nevertheless, depending on the correct interpretation of regulation 3, some features of Wikipedia might be considered to amount to a content recommender system or a forward/share functionality. That may mean that Wikipedia amounts to a Category 1 service within the meaning of the regulation. That is so even though these features are, for the most part, not available to general users and are only available to a tiny subset (such as moderators), and even though it might be thought that none of them have the effect of causing viral dissemination in the way that can be a feature of some social media sites.
It is sufficient to mention just one such feature: the New Pages Feed. The overwhelming majority of Wikipedia users will never encounter the New Pages Feed. It is a moderation tool. It helps moderators identify new pages so that they can be reviewed for accuracy and clarity to limit the spread of poor-quality content. The New Pages Feed involves the operation of algorithms, including code to identify and present a list of newly created pages, code to filter or sort the list, code that predicts the quality of the page, and code that applies a “tag” to different types of page (eg pages that are just a few words long (“stubs”), or which are not linked to any other articles (“orphans”) or which have no sources (“no citations”)).
A question that then arises is whether the New Pages Feed is a “content recommender system” within the meaning of regulation 3(2). That may raise the following further questions:
Is it “used by the [first claimant]”? Or is it only used by moderators who work entirely independently of the first claimant?
Is it used in respect of the user-to-user part of Wikipedia? Or is it only used in respect of another part of Wikipedia which does not form a component of the user-to-user part?
Does it use “algorithms” within the meaning of regulation 3(2)? Or is regulation 3(2) only intended to cover algorithms that affect how content may be encountered by users generally, as opposed to volunteers who take on a moderation role?
Does the New Pages Feed affect “the way in which user-generated content of a user… may be encountered by other users”? Or does it merely provide a mechanism for such content to be moderated?
Neither party seeks a finding on these questions, nor on the ultimate question of whether Wikipedia meets the Category 1 conditions. It is common ground that Wikipedia might do so, and that it will be for Ofcom to determine whether it does do so, giving a Convention compliant reading and effect to the regulation, so far as that is possible.
For its part, the first claimant’s litigation position in these proceedings (notwithstanding its response to Ofcom on 16 May 2025) is that it is “likely” that it meets the Category 1 conditions (and, on that basis, it contends that it is a “victim” for the purposes of the Convention). Conversely, Ms Rowland acknowledges that the Secretary of State was advised that it was “possible” that Wikipedia would qualify as a Category 1 service, but says that “because there are a number of technical issues which will need to be addressed by Ofcom when applying the Regulations concerning functionalities, systems and user numbers” the issue is necessarily uncertain.
More broadly, the issues that Ofcom will need to address before determining whether Wikipedia is a Category 1 service include:
Identifying the number of people in the United Kingdom who access Wikipedia. That is not straightforward, because it is difficult to distinguish between people and bots, and because one person may use more than one device (and different people may use the same device), and because it is not necessary to log in to Wikipedia to be able to access it, and because the use of virtual private networks makes it difficult to identify whether a person who accesses Wikipedia is in the United Kingdom. However, Wikipedia’s estimate (based on a methodology that appears reasonable, and which has not been challenged) is that approximately 26 million individuals in the United Kingdom access Wikipedia each month. If they are all users within the meaning of section 227 of the Act, then that comfortably meets the 7 million figure in regulation 3(1)(b)(i).
Determining whether each person who accesses Wikipedia is a “user” within the meaning of the legislation. That assessment will involve a question as to whether administrators such as BLN are users. That issue arises because persons who work for the first claimant, including as volunteers, are not (when acting in the course of the first claimant’s business) users for the purposes of the Act: section 227(3)(c). Nor is any other person who provides a business service to the first claimant (when that person is acting in the course of the first claimant’s business): section 227(3)(d). The claimants are keen to stress that Wikipedia is regulated by its user base, that its editors and administrators are entirely independent of the first claimant and that this is of fundamental importance to its operating model. On the other hand, the first claimant has urged Ofcom carefully to assess this issue, and it has advanced a cogent basis on which Ofcom might conclude that moderators are not (when acting as such) “users”. If moderators are not users, then that may be of considerable significance. That is because when it comes to assessing content recommender systems and forward and share functionalities, only those systems/functionalities that operate on the user-to-user part of the service fall to be considered. If moderators are not users, and if a content recommender system or forward and share functionality is only available to moderators, then that system/functionality will be left out of account.
Assessing whether each of a number of different features of Wikipedia amount to a content recommender system or a forward and share functionality. These include individual Wikipedia pages (including the main page) and banners and templates, the Wikipedia search function (including search query autocompletion and ranking of results by relevance), the New Pages Feed, a recent changes feed, a content translation tool, a QR code generator, a short-url generator, a “cite this page” function, and a “newcomer tasks” feature.
Submissions
Ground 1: Compliance with paragraph 1(5) of schedule 11
Rupert Paines, for the claimants, submits that the Secretary of State failed to consider the full likely impact of the two critical features (content recommender systems and forward/share functionalities) on viral dissemination, as required by paragraph 1(5) of schedule 11. On analysis of Ofcom’s advice and the Secretary of State’s acceptance of that advice, there was an erroneous assumption that these features are always integral to an online service, and that they inter-operate to cause viral dissemination. The Secretary of State’s consideration of those features was therefore based on, and limited to, an assumption that they are integral to the service that they inter-operate, disregarding instances (such as in the case of Wikipedia) where they are ancillary, where they operate independently of one another, and where they do not give rise to any risk of viral dissemination. That omission meant the Secretary of State did not comply with his duty under paragraph 1(5) of schedule 11.
Cecilia Ivimy KC, for the defendant, submits that the obligation under paragraph 1(5) of schedule 11 does not require a granular analysis to consider the different ways in which these features might operate, as posited by the claimants. Nor was the Secretary of State required to consider every possible interaction of functionalities and user numbers. He was entitled to rely on Ofcom’s research and advice. The Secretary of State did consider the likely impact of user numbers and functionalities on viral dissemination, and that was all that was required by paragraph 1(5) of schedule 11.
Ground 2: Rationality
Mr Paines submits that Regulation 3 captures services with content recommender systems and forward/share functionalities even if those functionalities are not integral to the service and thus do not give rise to a risk of viral dissemination. That involves a failure both of reasoning and investigation which has resulted in overbroad criteria. It was also illogical not to take account of the time spent by users accessing the service (which, for Wikipedia is around 18 minutes per month, compared to 32 hours per month for Meta sites such as Facebook and Instagram, and Alphabet sites such as YouTube and Google search) when calculating the number of monthly users. For these reasons, the decision to make the regulation was irrational.
Ms Ivimy argues that the regulations have a logical underpinning that derives from Ofcom’s research and advice. In the light of that research and advice it was reasonable and logical to use content recommender systems and forward/share functionalities, in conjunction with a high minimum user number threshold, to capture services that are likely to involve viral dissemination. Even if the conditions are imperfect, it is permissible to adopt broad regulatory measures in what is a highly complex and technical field.
Ground 3: Compatibility with articles 8, 10 and 11 of the Convention
Mr Paines submits that the user verification duties would fundamentally disrupt Wikipedia’s collaborative model. They would force Wikipedia to either restrict access or implement unworkable changes which are not compatible with the way in which Wikipedia operates (for example, in relation to ensuring that users can opt not to encounter content from (or have their content edited by) non-verified users). This inevitably amounts to an unjustified interference with the first claimant’s rights under articles 10 and 11 of the Convention, and the second claimant’s rights under articles 8, 10 and 11.
Ms Ivimy says that the claimants have not established that they are victims of a breach of the Convention so as to give them sufficient standing to bring the claim. The claim is, at this stage, hypothetical and the question of whether there is, ultimately, a breach of Convention rights will depend on the correct interpretation of the regulation (taking account of the interpretative obligation under section 3 of the Human Rights Act 1998) and the regulatory decisions made by Ofcom. It is not appropriate to second-guess how the statutory and regulatory process will play out.
Ground 4: Breach of articles 14 of the Convention/Irrationality
Mr Paines submits that Regulation 3 unjustifiably treats Wikipedia, a non-profit, collaborative encyclopaedia, in the same way as major profit-driven social media companies, despite their obvious and significant differences. This blanket approach constitutes a breach of the prohibition of discrimination contrary to article 14 of the Convention, which provision requires that relatively different organisations are treated differently. There is no objective or reasonable justification for the Secretary of State’s decision to treat Wikipedia in the same way as a major social media business.
Ms Ivimy responds that this ground of claim is entirely hypothetical. It is not yet known if the first claimant will fall within Category 1 and thus whether it will be treated in the same way as social media companies. In any event, the services that fall within the scope of regulation 3 differ in many different ways, and this was always understood and expected. That is an inevitable, and justified, consequence of the policy aim that underpins regulation 3, namely identifying conditions that capture, in a manner that is objective and workable, services that are likely to involve viral dissemination.
Ground (1): Did the defendant breach paragraph 1(5) of schedule 11 of the Act?
Paragraph 1(5) of schedule 11 of the Act required the Secretary of State, when making the regulations, to take account of the likely impact of the number of users of a service, and its functionalities, on viral dissemination.
The Secretary of State took account of the impact of the number of users. He accepted Ofcom’s advice as to the thresholds to be applied. A separate issue arises in relation to how numbers of users should be calculated, but that is relevant to ground (2). The real issue under ground (1) concerns functionalities.
Ofcom considers that content recommender systems do not, strictly, fall within the statutory definition of a functionality. Whether it is right or wrong about that, for these purposes a content recommender system can be treated as a functionality. That is because even if it does not fall within the statutory definition, both parties agree that its impact on the dissemination of content is a relevant factor that the Secretary of State was obliged to take into account at common law. Moreover, Ofcom treated content recommender systems, for these purposes, as though they are functionalities. I would not, therefore, dismiss this ground of challenge, so far as it concerns content recommender systems, on the basis that such systems are not functionalities.
Ofcom, as the relevant and expert regulator, was under a statutory obligation to conduct research into how easily, quickly and widely user-generated content is disseminated by user-to-user services, and the functionalities of the user-to-user part of such services: paragraph 2(2) of schedule 11. In conducting that research, a significant focus was the impact of different functionalities on viral dissemination. It then provided advice that was based on its research, as required by paragraph 2(5) of schedule 11. The Secretary of State was required to, and did, take account of that advice.
By the time of the first Ministerial Submission to the (new) Secretary of State, there had been extensive discussion of the issues at both Ministerial and official level and there had been a degree of lobbying. Ms Rowland’s evidence is that those discussions fed into the briefings that were given to the Secretary of State. That evidence is consistent with the contemporaneous documents. The advice that Ofcom had provided was well understood. There was an appreciation that if the advice was accepted then the consequences for which services would fall within the scope of Category 1 were not entirely clear. Some services were likely to be in scope; others, including Wikipedia, were “possibilities” and were considered to be “outliers”. There was also an appreciation of the benefits and risks of not accepting Ofcom’s advice and/or seeking further advice.
The Submission of 18 July 2024 told the Secretary of State that the issue was controversial among stakeholders and Parliamentarians. He was told that he was required to take account of the likely impact of a service’s functionalities on viral dissemination. The uncontested evidence of Ms Rowland shows that the Secretary of State was closely engaged with the relevant issues. He did not simply approve the recommendation of Ofcom. The Minister had a meeting to express concern about the impact of the advice on “outlier” services, and options were then considered to exclude those services, including Wikipedia. Those options included conducting further research into viral dissemination and to understand more about the operation of content recommender systems. That was presented, in detail, in the submission of 31 July 2024. Again, the Secretary of State did not simply approve the recommendation to accept Ofcom’s advice. A further meeting took place, attended by the Secretary of State, on 12 August 2024. At that meeting, the Secretary of State considered the impact of Ofcom’s advice on online encyclopaedias and was content not to seek further advice on this issue.
The Secretary of State thus considered and accepted Ofcom’s advice. He therefore considered and accepted the advice that, for large services, content recommender systems and a functionality to forward and share user-generated content (but not so much other functionalities) impacted on viral dissemination and appreciated that this advice was being given at a high level of generality. It follows that he took account of the likely impact of a service’s functionalities on viral dissemination. It follows that he complied with paragraph 1(5) of schedule 11 so far as functionalities are concerned.
The claimants’ key argument is based on a literal reading of paragraphs 3.23 and 3.24 of Ofcom’s advice, which were faithfully reflected in the material put before the Secretary of State. On the basis of a literal reading of those paragraphs, the claimants say that Ofcom advised that the two key functionalities that they identified were “alone sufficient” for viral dissemination. It follows, say the claimants, that Ofcom necessarily assumed that the two functionalities would be integral to the service and would operate and interact to disseminate content to users. However, that is only the case where the content recommender system supplies content to the majority of users. Ofcom’s logic does not apply to a feature like the New Pages Feed where the content recommender system operates in a quite different manner to assist a small group of moderators to ensure the quality of user-generated content. They say that the Secretary of State never addressed his mind to this issue. Instead, he necessarily restricted himself to consideration of the impact of content recommender systems and forward and share functions which are integral to a service, and which operate in conjunction with each other.
The claimants may well be right that there are instances (and the New Pages Feed may well be a good example) where a feature that, on one view, might amount to a content recommender system does not promote viral dissemination, and where it does not operate in conjunction with a forward and share functionality. They are also no doubt correct that the Secretary of State did not consider all the different ways in which content recommender systems (or forward and share functionalities) might operate, and the impact which each different mode of operation might have on content dissemination.
This does not, however, amount to a breach of paragraph 1(5). The claimants’ argument does not take account of the nature of Ofcom’s advice, the nature of the obligation imposed by paragraph 1(5), and the evidence about the Secretary of State’s consideration of the issues.
As to the nature of the advice, Mr Paines is correct that paragraphs 3.23 and 3.24, when read in isolation, appear to indicate that where a service has a considerable number of users, a content recommender system and a forward/share functionality are sufficient (irrespective of how they operate in practice) to increase the likelihood of viral dissemination. However, those paragraphs need to be read in the context of the advice as a whole. When that is done, they cannot properly be understood in such a literal and inflexible sense:
The advice included a detailed account of the research that had been undertaken.
That research made it clear that there were no existing independent assessments of the functionalities provided by different online services, that there was no widely adopted definitions for functionalities, and that online services regularly changed their functionalities. Ofcom had to undertake “an indicative assessment.”
There are a vast number of online user-to-user services, across many different sectors (including social media, information-sharing, pornography, marketplaces, listing services, video-sharing, retail, audio streaming, gaming, private communications and file sharing). Ofcom’s research did not purport to cover, comprehensively, all different permutations.
Ofcom recognised that the policies of an individual service could limit viral dissemination, irrespective of the presence of any particular functionalities.
Ofcom stressed the importance of user numbers, observing that even where the functionalities and characteristics it was discussing were “core to the service”, viral dissemination would be less likely where the user base is smaller (thereby indicating that Ofcom was, more generally, not limiting its consideration to functionalities that were “core to the service”).
The summary of the research findings “indicate[d]” that content recommender systems and a forward/share functionality were the features of a service that were “most relevant” to the viral dissemination of content, that content recommender systems were “typically” relied on to amplify content, and a forward/share function “facilitates” the sharing of content.
The advice made it clear that it was based on that research.
The summary of the advice said that the research found that these are the features that are “most relevant” to viral dissemination (not that they necessarily result in that consequence, irrespective of how they operate).
Ofcom well understood that the role of its research and advice was to inform the Secretary of State’s assessment as to the likely impact of functionalities on content dissemination.
Ofcom made clear that in providing its advice on categorisation it was exercising a “regulatory judgement based on [its] general duties and functions.”
As to the nature of the obligation imposed by paragraph 1(5), it is pitched at a high level of generality. The Secretary of State is required to assess the “likely” impact of functionalities (generally) on content dissemination. That is an assessment that has to be made across the entire universe of online regulated user-to-user services. There are a vast number of such services operating in many different sectors. Nothing in the statutory language implies that the Secretary of State is required to consider each different functionality of each different service in each different sector, or that he is required to determine the precise impact of any particular functionality on content dissemination, or that he is required to consider the many different ways in which individual functionalities might operate.
The core of the claimants’ case is that the Secretary of State was obliged to consider the different ways in which functionalities might operate where they are not integral, or core, to a service, and to recognise that in such cases they are unlikely to result in viral dissemination. I do not agree. Nothing in paragraph 1(5) requires the Secretary of State to draw that distinction or undertake that level of analysis. The obligation is pitched at an altogether higher level. The advice that the Secretary of State was given was that the operation of content recommender systems was often opaque, and that they regularly change. No doubt, they operate in many different and changing ways. The claimants focus on a distinction that might assist their particular case, but many other distinctions could be considered. There may be possible distinctions between core and non-core functionalities; or between integral and non-integral functionalities; or between functionalities that operate across the whole user-base and those that impact on only a limited sub-set of users; or between those that are relatively unconstrained and those that are constrained by the operation of policies; or between those that are intended to amplify content and those that are intended to restrict content; or between those that are intended for moderation purposes and those that are not; or between those that are operated in the not-for-profit sector and those that are operated by large profit making organisations where viral dissemination is important to the business model. It was not necessary (or practical, or possible) for the Secretary of State to work through all the different possible permutations. Moreover, the distinction that the claimants seek to draw is not hard-edged. They do not propose any test that would determine whether a content recommender system is “integral” or “non-integral” to a service. It is not easy to formulate such a test. The advice that the Secretary of State was given was that further advice would be required to undertake a more granular analysis, but that until the first regulations had been made it was unclear that Ofcom had a statutory power to provide such advice. It is thus unclear how the Secretary of State could, in practice, have distinguished between the effect on viral dissemination of integral and non-integral functionalities, or how he could have undertaken anything other than the high level and general assessment that is required by paragraph 1(5).
As to the evidence of the Secretary of State’s consideration of the issues, although he did not, and could not, undertake the type of analysis that the claimants suggest, he was well aware of the complexities, and of the fact that Ofcom’s advice might produce some results (possibly in both directions) that did not reflect the underlying policy intent. Hence, the focus on “outliers” and the potential that a service such as Wikipedia, which officials (and seemingly Ofcom) had not originally contemplated falling with Category 1 might do so. All of that was considered.
Accordingly, the Secretary of State therefore did, in making regulation 3, take into account the likely impact of a service’s functionalities on viral dissemination. There was no breach of paragraph 1(5) of schedule 11.
For these reasons, while I grant permission to claim judicial review on ground (1), this ground of claim is dismissed.
Ground (2): Was the decision to make regulation 3 irrational?
The decision to make regulation 3 was irrational if it was a decision which no reasonable Secretary of State could have made. No reasonable Secretary of State could make a regulation if the basis for doing so involved demonstrably flawed reasoning which robbed the regulation of a sound logical basis: R (Law Society) v Lord Chancellor [2018] EWHC 2094 (Admin); [2019] 1 WLR 1659 per Carr J at [98].
There is a perfectly cogent argument that it would be irrational to decide that Wikipedia has a content recommender system and a forward/share functionality that causes viral dissemination of its content. That is not, however, the decision that the Secretary of State made; and he was not required to address that issue.
The decision that the Secretary of State was required to make was mandated by paragraph 1(1) of schedule 11 to the Act. It was to specify Category 1 threshold conditions for the user-to-user part of regulated user-to-user services relating to the number of users of that part of the service and the functionalities of that part of the service and any other characteristics of that part of the service or factors relating to that part of the service that the Secretary of State considered relevant. The regulations were required to specify the ways in which the threshold conditions may be met and to specify at least one condition about number of users or functionality: paragraph 1(4). The Secretary of State was required to take account of the likely impact of the number of users and the functionalities on viral dissemination: paragraph 1(5). The Secretary of State did so, which is why ground (1) of this claim is dismissed. The Secretary of State was also (at least implicitly) required to take account of Ofcom’s advice: paragraph 2(5), 2(7)(b), 2(8), 2(11) of schedule 11.
The decision as to the criteria to be set under paragraph 1(1) fell to be made in a highly complex and technical policy area which requires difficult evaluative regulatory assessments. It was impossible to know, on the information available, precisely what the impact would be on individual services. The regulation would have no immediate practical impact on any individual service. There was an opportunity later to amend the regulation, or to exempt particular types of service, if it turned out that the regulation had undesirable consequences. These are all important aspects of the relevant context.
I agree with Mr Paines that the technical context does not immunise the decision from the court’s review. Nor does the fact that the Secretary of State agreed with the recommendation of his officials, formulated after months of consideration. Nor does the fact that the recommendation was to follow statutory advice that was provided by the expert regulator. Nor does the fact that that advice was itself based on statutory research undertaken by the regulator. Nor does the fact that the draft regulations were approved by a resolution of both Houses of Parliament. These are, however, all further important aspects of the relevant context. They inform the degree of circumspection to be applied before concluding that the decision was irrational. If there was some basic logical flaw in the reasoning which robbed the decision of its logic, then it was a flaw that seemingly evaded the expert regulator, experienced policy officials, the Minister and the Secretary of State (and the previous Secretary of State).
Ofcom had put forward closely researched and reasoned advice as to criteria that would, broadly, capture services that give rise to viral dissemination. It was well understood that the proposed criteria might capture some services that do not give rise to viral dissemination and might miss some other services that do give rise to viral dissemination. Once that is recognised, the claimants have not identified any basic flaw in the logic or reasoning that Ofcom applied, and which officials approved, and the Secretary of State accepted. It is easy to see why, in many cases, a content recommender system in association with forward/share functionality is likely, if the user-base is sufficiently large, to give rise to viral dissemination. It is also easy to see why the absence of a content recommender system and a forward/share functionality is likely to limit the potential for viral dissemination. That easy intuition accords with the research that Ofcom conducted.
The real problem is that the criteria are not perfect. They may capture some systems (possibly including Wikipedia) where there may be no real risk of viral dissemination. However, neither Ofcom nor the Secretary of State were acting under a misapprehension that the criteria were perfect in the sense that they would capture all services that give rise to viral dissemination and would not capture any services that do not do so. If there were some obvious alternative criteria that were perfect in that sense (or which objectively more precisely covered the relevant field) then it may have been irrational to accept the officials’ recommendation to follow Ofcom’s advice. However, even now, neither the claimants nor anyone else have suggested a worked-through better alternative formulation.
To the extent that the claimants suggest that the criteria could be limited to content recommender systems and forward/share functionality that are “integral” to the system, they have not identified any research or evidence that this would lead to better alignment with the policy intention (even assuming that it is possible sufficiently to define what is meant by an “integral” characteristic or functionality). Moreover, the Secretary of State was entitled to accept the warnings of his officials that to depart from Ofcom’s evidence and research-based advice would risk creating loopholes, and that it was not practical, at that stage, to seek further advice.
To the extent that the claimants suggest that user-numbers should be defined in a way that takes account of the time that users spend on a service, this suggestion appears to owe more to the particular characteristics of Wikipedia (where users do not typically spend a huge amount of time browsing the system, in comparison with the time typically spent by users on some social media sites) rather than clear evidence as to the relationship between time spent on a service and viral dissemination. Ofcom had not suggested this as a relevant metric and nor had officials. Ms Rowland’s evidence is that officials did consider whether the definition of a content recommender system could be limited by reference to time spent viewing content generated by such a system but concluded that there would be substantial workability issues with that option as it would be very difficult to measure. In the absence of any clear independent evidence that such a metric could easily be devised, that is not an unreasonable or illogical conclusion.
The fact that the criteria may capture some services that do not give rise to viral dissemination does not mean that they are irrational. In the Law Society case, Carr J observed, at [113]:
“A policy-maker may reasonably decide that the disadvantages of a finely tuned solution to a problem outweigh its advantages and that a broader measure is preferable, even if the broader measure is both over- and under- inclusive in that it catches some cases in which there is no or no significant problem and fails to catch some cases in which the problem occurs.”
I accept a submission that was advanced by Mr Paines that bright line rules (which may by their nature be both over- and under- inclusive) are more usually associated with contexts where a very large number of individual decisions need to be made, and tailored decision making must necessarily yield to the desirability of certainty and practicality. Well-known examples include social benefits and immigration systems: Pantellerisco v Secretary of State for Work and Pensions [2021] EWCA Civ 1454 per Underhill LJ at [59], R (Refugee and Migrant Forum of Essex and London) v Secretary of State for the Home Department [2024] EWHC 1374 (Admin) per Cavanagh J at [192] – [193]. I also accept that this context is different. There are likely to be relatively few Category 1 services (the evidence suggests less than 20). A legislative choice could have been made in the Act to enable the regulator to make tailored individualised decisions as to which services should fall within Category 1. However, that was not the choice made by Parliament. Paragraph 1(1) of schedule 11 requires the setting of boundaries by reference to user numbers, functionalities and characteristics, rather than permitting a decision-maker a free hand to determine on a case-by-case basis which services should fall within Category 1. Whether or not that choice was made because of the need for the system to operate fairly across many different types of online service, or for it to be future-proof in an area which is rapidly developing, or for it to cohere with categories 2A and 2B (which are likely to involve many more services), or for it to provide an objective and workable regulatory framework in a complex technical area, is beside the point. The fact is that this was the legislative choice made by Parliament. As Baroness Jones put it, “I recognise that our hands are largely tied by the constraints of the Act.” The Secretary of State was obliged to make regulations accordingly.
A separate issue arises as a result of the comment, in the Ministerial Submissions of 18 and 31 July 2024, that Wikipedia has content of democratic importance and journalistic content. That does not, in itself, provide a rational basis for drafting the Category 1 thresholds in a manner that captures Wikipedia. The mere fact that it contains content of democratic importance and journalistic content does not mean that there is a risk of viral dissemination (which is what the Category 1 threshold is intended to capture). However, nothing in the Submission, or in the decision-making process, suggests that the Secretary of State thought otherwise. The reference to content of democratic importance and journalistic content arose in the context of Ministerial concern about “outlier” services, including online encyclopaedias, which it had not originally been thought would be captured by Ofcom’s criteria, and consideration being given to an amendment to carve out certain services. The point was that if Wikipedia was otherwise captured by Ofcom’s criteria, then, in deciding whether to consider some form of carve out, it was relevant to take account of the fact that it contained content of democratic importance and journalistic content. Given that such content is of high importance to the “public discourse” policy consideration that underpins Category 1, that was a legitimate observation to make.
For these reasons, while I grant permission to claim judicial review on ground (2), this ground of claim is also dismissed.
Ground (3): Incompatibly with articles 8, 10 and 11 of the Convention
It is, generally, unlawful for a public authority to act in a way that is incompatible with a Convention right: section 6(1) of the Human Rights Act 1998. That can, in principle, include making a regulation that is incompatible with a Convention right. Section 6(1) does not apply if, as the result of primary legislation, the public authority could not have acted differently: section 6(2)(a). Nor does it apply if the public authority was acting so as to give effect to a provision made under primary legislation which cannot be read or given effect in a way which is compatible with Convention rights: section 6(2)(b). The Secretary of State does not rely on section 6(2).
A person who claims that a public authority has acted in a way which is unlawful by section 6(1) may bring proceedings under the 1998 Act: section 7(1)(a). They may only do so if they are a victim of the unlawful act: section 7(1). Moreover, a claimant only has a sufficient interest to bring a claim for judicial review in respect of such an act if they are a victim of that act: section 7(3).
A person is a victim of an act that is unlawful by reason of section 6(1) only if they would be a victim for the purposes of article 34 of the Convention if they brought proceedings in the European Court of Human Rights: section 7(7).
The Grand Chamber of the European Court of Human Rights has recently restated the test to establish victim status within the meaning of article 34 – Verein Klimaseniorinnen Schweiz v Switzerland (2024) 79 EHRR 1 at [465] – [470]:
“465. In order to fall into the category of direct victims, the applicant must be able to show that he or she was “directly affected” by the measure complained of... This implies that the applicant has been personally and actually affected by the alleged violation of the Convention, which is normally the result of… acts… of State authorities… allegedly infringing the applicant’s Convention rights…
466. However, this does not necessarily mean that the applicant needed to be personally targeted by the act or omission complained of. What is important is that the impugned conduct personally and directly affected him or her…
469. Two types of potential victim status may be found in the case-law… The first type concerns persons who claim to be presently affected by a particular general legislative measure. The Court has specified that it may accept the existence of victim status where applicants contend that a law violates their rights, in the absence of an individual measure of implementation, if they belong to a class of people who risk being directly affected by the legislation, or if they are required either to modify their conduct or risk being prosecuted (see Tănase v Moldova…)…
470. The second type concerns persons who argue that they may be affected at some future point in time. The Court has made clear that the exercise of the right of individual petition cannot be used to prevent a potential violation of the Convention and that, in theory, the Court cannot examine a violation other than a posteriori, once that violation has occurred. It is only in highly exceptional circumstances that an applicant may nevertheless claim to be a victim of a violation of the Convention owing to the risk of a future violation… In general, the relevant test to examine the existence of such victim status is that the applicant must produce reasonable and convincing evidence of the likelihood that a violation affecting him or her personally will occur; mere suspicion or conjecture being insufficient in this regard…”
The claimants do not positively contend that Wikipedia is captured by regulation 3. It follows that they are unable to show that they fall within the first category of victim. The claimants say that it may turn out, in due course, that Wikipedia is captured by regulation 3. However, Verein Klimaseniorinnen Schweiz shows that it is only in highly exceptional circumstances that victim status arises because of a risk of a future violation of the Convention. The claimants must produce reasonable and convincing evidence of the likelihood that a violation of their Convention rights will occur.
This approach to victim status is apt where the events which will give rise to a breach of the Convention have not yet occurred but will, on the evidence, likely occur in the future. In Asselbourg and other v Luxembourg (application 29121/95, 29 June 1999) the contingency was the likelihood of future damage sufficient to cause a violation of Convention rights based on the discharge of air-polluting waste. In Senator Lines GmbH v Austria and others (2004) 39 EHRR SE3 (application 56672/00, 10 March 2024) the contingency was the risk that a fine would be upheld. In Verein Klimaseniorinnen Schweiz, thecontingency was the likelihood that the individuals represented by the applicant would be personally and directly affected by the state’s failure to take reasonable measures to reduce harm due to anthropogenic climate change.
The nature of the contingency here is quite different. The real contingency is not whether Ofcom will decide that Wikipedia is a Category 1 service. Ofcom does not have a choice, or any discretion, about that. It must apply regulation 3 (interpreted in accordance with section 3 of the 1998 Act). If regulation 3 (thus interpreted) covers Wikipedia, then it is a Category 1 service. If it does not, it is not.
The real contingency is therefore the correct interpretation of regulation 3. That is not a matter that depends on evidence as to future events. It is, in principle, an issue that is capable of being determined by a court now. It would have been open to the claimants to ask the court to do that, and to advance a positive argument that regulation 3 applies to Wikipedia. They have chosen not to do that, preferring to advance an argument on that issue in the context of Ofcom’s decision-making process. However, that choice means that they are unable, in these proceedings, to demonstrate that they are a victim. It is not open to them to argue that they are a victim because Ofcom, or a court, may interpret regulation 3 in a particular way in the future – that is not the type of highly exceptional circumstance in which it is appropriate to ascribe victim status on the basis of a risk of future violation.
Mr Paines correctly pointed out that the interpretation of regulation 3 is a question of law for the court. He complained that the Secretary of State had not addressed this issue and submitted that if the Secretary of State’s case is that regulation 3, properly construed, excludes a service such as Wikipedia then he should have said so. The same, however, applies to the claimants. If it is their case that Wikipedia falls within the scope of regulation 3 then, on Mr Paines’ logic, they should have said so. The fact is that both parties have chosen to litigate this claim in a way that seeks to avoid the critical central issue.
Mr Paines submits that it would be open to the court to reach a definitive interpretation of regulation 3, even though that is not a course that either party has invited the court to take. To the extent that was an implicit invitation to embark on such an exercise, I decline to do so. It is a matter of considerable importance, and it would be inappropriate to reach a determination without hearing full argument from all parties. Moreover, there is sense in the approach that both parties have taken of leaving it to Ofcom to undertake its regulatory functions and to decide for itself, in the first instance, how regulation 3 should be interpreted. The problem for the claimants is that a consequence of that approach is that they are unable to demonstrate that they are victims for the purpose of section 7 of the 1998 Act.
Even if it were apt to apply the second limb of the Verein Klimaseniorinnen Schweiz test in the present case, the claimants have not shown that it is satisfied. In particular, I do not accept that it is “likely” that Wikipedia is a Category 1 service. The furthest that Ofcom has gone is to say that it might be. The Secretary of State’s officials considered only that it was a “possibility” (whereas other services were “likely” to qualify). The first claimant has, in effect, invited Ofcom to conclude that it does not qualify, and it has advanced cogent arguments in support of that invitation. The issue does not depend simply on a natural construction of the words of regulation 3, because the regulation must be read, and given effect, in a way that is compatible with Convention rights. There is, arguably, more than one way in which regulation 3 might be interpreted, compatibly with Convention rights, so as to produce an outcome that does not capture Wikipedia. That includes the approach to be taken to “users” (including whether administrators or moderators are properly to be regarded as “users” when they are acting as such); the approach to be taken to “content recommender systems” (including whether they are to be taken as systems that recommend content to users generally (as opposed to administrators/moderators)); and the approach to “functionality for users to forward or share regulated user-generated content” (including whether that is to be taken as functionality that applies to users generally, or whether it applies to a functionality that is only available to a small subset of the user base). The claimants have not shown (and have not sought to show) that regulation 3 is incapable of being interpreted in a manner that is compatible with Convention rights.
The claimants rely on Tănase v Moldova (2011) 53 EHRR 22 and Michaud v France (2014) 59 EHRR 9 and say that they are members of a class of people who risk being directly affected by the legislation. However, these are cases which fall within the first category of victim described in Verein Klimaseniorinnen Schweiz at [469] (where Tănase is cited) and require the claimants to show that they are presently affected by the legislation. They have not done that.
Even if it turns out that Wikipedia meets the criteria of regulation 3, as it is currently drafted, it does not necessarily follow that this will lead to a breach of Convention rights. Mr Bradley-Schmieg and BLN advance a powerful case that the application of Category 1 duties to Wikipedia is incompatible with (at least) article 10 of the Convention. If they are right about that, then the Secretary of State may be required to take further action (whether by amending the regulation or by exempting a class of service that includes Wikipedia).
I do not, however, consider that it is inevitable that the application of Category 1 status to Wikipedia would necessarily result in a breach of Convention rights. That may depend on how the Code of Practice (compliance with which is deemed to achieve compliance with the statutory duties) operates. One of Wikipedia’s primary concerns is the requirement to enable users to choose only to encounter content from users whose identity has been verified. I accept that this runs completely counter to Wikipedia’s operating model (which has, on the evidence, been shown to be effective in promoting freedom of expression whilst promoting a high quality of content). There may, however, be ways to accommodate the requirement without causing undue damage to Wikipedia’s operations. It may, for example, be possible to ensure that users who make the requisite choice are only able to access pages where every editor who has contributed to the live content on the page has verified their identity. It is not obvious that this would be unduly difficult to achieve. It would mean that such users would only be able to access a small subset of Wikipedia’s content, but that would be their free, autonomous choice. It may well be that only very few people would make that choice, and that might then raise a question as to the proportionality of the entire exercise. This type of approach might, though, be sufficient to address section 15(10)(b) of the Act (it would also be necessary separately to address section 15(10)(a)).
These are, however, all matters that are much better addressed further down the road. The parties implicitly recognise that, by declining to litigate, at this stage, the fundamental issue of the ambit of regulation 3. Further, I accept Ms Ivimy’s submission that a decision as to whether the claimants’ Convention rights are breached will require an intense analysis of the facts which cannot be carried out in the abstract or by reference to hypothetical scenarios: Verein Klimaseniorinnen Schweiz at [460], Imperium Trustees (Jersey) Limited v Jersey Competent Authority [2025] UKPC 28; [2025] 1 WLR 3225 per Lord Lloyd-Jones (giving the judgment of the Board) at [78].
For these reasons, the claimants have not established that they are victims, or that they have standing to bring ground (3) of this challenge. Accordingly, I refuse permission to advance ground (3).
Ground (4): Irrationality/Incompatibly with article 14 of the Convention
The reasoning in respect of victim status and standing under ground (3) applies equally to ground (4) so far as that is based on article 14 of the Convention. For that reason, I refuse permission on ground (4). There are, however, additional difficulties with the claimants’ case on ground (4).
Article 14 of the Convention gives a right to the enjoyment of other Convention rights without discrimination on a specified ground. That can include an obligation to treat differently persons whose situations are significantly and relevantly different (or, otherwise, to provide an objective and reasonable justification for treating them in the same way): Thlimmenos v Greece (2000) 31 EHRR 12 at [44], R (DA) v Secretary of State for Work and Pensions [2019] UKSC 21; [2019] 1 WLR 329 per Baroness Hale at [136].
The claimants’ complaint under ground (4) (whether presented as an irrationality challenge or a breach of article 14) is that the Secretary of State should have drawn a distinction between non-profit making charities such as the first claimant and major highly profitable technology companies (such as Alphabet, Meta and X) that run large social media and chat platforms (such as Facebook, Instagram, YouTube and X). The premiss for the complaint is that the legislation was intended to target “profit making social media companies”. The premiss is, however, misplaced. Although it is possible to point to instances of political rhetoric where reference has been made to the profits made by companies in Silicon Valley, nothing in the legislation (or in the permissible aids to its interpretation) show that its focus was limited to any particular type of company. Rather, it is clear that it is intended broadly to capture and regulate online companies that give rise to different types of risk across a wide range of online providers of different types.
There are some forms of risk that may be inherently less acute in the case of a non-profit making charity. For example, if such a charity does not carry adverts, then fraudulent advertising will not be an issue. But there is no evidence that the broad risks to public discourse that are intended to be captured by a Category 1 designation are limited to profit-making companies. On the evidence, the fact that the first claimant does not make a profit does not distinguish it from other online providers so far as concerns its potential to influence public discourse.
For these reasons, I refuse permission to claim judicial review on ground (4).
Outcome
I refuse permission to claim judicial review on grounds (3) and (4) because the claimants do not have standing to bring a claim under the Human Rights Act 1998 and because those grounds of claim do not (as matters stand) have arguable merit.
I grant permission to claim judicial review on grounds (1) and (2), but I dismiss the claim. The claimants have not shown that the decision to make regulation 3 is flawed on any of the grounds that they have advanced.
I stress that this does not give Ofcom and the Secretary of State a green light to implement a regime that would significantly impede Wikipedia’s operations. If they were to do so, that would have to be justified as proportionate if it were not to amount to a breach of the right to freedom of expression under article 10 of the Convention (and, potentially, a breach also of articles 8 and 11). It is, however, premature to rule on that now. Neither party has sought a ruling as to whether Wikipedia is a Category 1 service. Both parties say that decision must, for the moment, be left to Ofcom. If Ofcom decides that Wikipedia is not a Category 1 service, then no further issue will arise.
Ofcom’s decision as to which services fall within Category 1 is a public law decision which is potentially amenable to the court’s review on grounds of public law error (including incompatibility with a Convention right). So, if Ofcom impermissibly concludes that Wikipedia is a Category 1 service, the claimants have a remedy by way of judicial review.
If Ofcom permissibly determines that Wikipedia is a Category 1 service, and if the practical effect of that is that Wikipedia cannot continue to operate, the Secretary of State may be obliged to consider whether to amend the regulations or to exempt categories of service from the Act. In doing so, he would have to act compatibly with the Convention. Any failure to do so could also be subject to further challenge. Such a challenge would not be prevented by the outcome of this claim.