Today (25th February 2025) the Independent Publisher Alliance submitted its response to the UK government's consultation on copyright and artificial Intelligence. Deepest thanks to Alliance Board Member David Buttle for your efforts in allowing us a coherent and effective argument on behalf of our members.
The Full submission here....
The Independent Publishers Alliance (‘Alliance’ hereafter) is a dynamic non-profit organisation committed to championing and advocating for the UK’s independent publishers. The Alliance represents 50 businesses, based all over the country, that together employ over 1500 media professionals in the UK and serve audiences in the tens of millions. The Alliance welcomes the opportunity to respond to this consultation. Clear and enforceable intellectual property law is a fundamental cornerstone of the success of our members. More widely, the UK’s world-leading creative sector has flourished in part as a result of the gold-standard IP regime in this jurisdiction. Because creators, and creative businesses, have been able to reap the economic reward for their talent and work, they have been incentivised to innovate which in turn has delivered huge economic benefits to the country. Today the creative sectors are worth £125Bn to the UK economy. We are deeply concerned that the government’s proposals for a weakening of our copyright law will undermine this success story, harming the creative industries, independent publishers and, as a consequence, the UK more widely. It risks doing so for no concrete benefits but instead, speculative gains from the AI sector; one in which the UK is unlikely to supplant the US or China’s leadership positions. The position of the Alliance is not anti-AI. We can see the huge potential of this technology and many of our members are already deploying it, judiciously, in both the operation of our businesses and to assist in the creation of content for our audiences. However, we believe that the government’s support for the AI sector cannot come at the expense of other industries. The government’s proposals come at a time when existing IP law is being flagrantly violated at vast scale for the benefit of foreign-owned businesses and their investors. This is already causing substantial economic damage to the UK’s independent publishing sector. Rather than protecting businesses whose core product is being stolen, repackaged and monetised without payment, the government is supporting those perpetrating this theft and proposing to legalise it under certain circumstances. The ability of IP holders - and particularly publishers - to protect their works when used by AI systems is becoming more important each day, as consumers adopt AI tools and these are used to access and retrieve information on their behalf. Publishers will need to adapt their strategies to this new platform environment. Existing monetisation via advertising and subscriptions will be diminished and it will need to be replaced by revenue secured via non-human access to owned-and-operated properties. This means licensing. Rather than focus on policy that would foster the development of a vibrant licensing market, the government’s proposals would undermine its formation.
The Alliance’s response to the questions posed by the consultation follows below, organised by section of the consultation document.
Policy options
The government’s proposal to pursue option 3 - a text and datamining (TDM) copyright exception with a rights holder opt-out - is predicated on a number of assumptions for which there is no supporting evidence and with which the Alliance does not agree.
Firstly, the consultation argues that creating a more permissive IP regime will result in the training of foundation models taking place in the UK. No market assessment has been published as the basis for this claim and the existing practice of AI developers suggests it is unlikely to be the case. The most permissive TDM regimes in the world exist in Japan and Singapore. No major models have been trained in either of these jurisdictions. This proposal in effect brings the UK's regime in line with the EU, where very little AI model development is taking place now. If the policy objective is for training to take place in the UK, this proposal seems highly unlikely to achieve this.
The consultation is also predicated on the idea that there is ambiguity in the application of UK copyright law to the use of protected content by artificial intelligence developers. Option 0 is dismissed for this reason. We contend that this is simply not the case. The UK’s copyright regime is clear. The problem that needs to be resolved is not a lack of clarity but that the law is not being respected: Content is being taken without authorisation or payment and publishers have not been approached for a licence for training. Existing IP law needs to be applied and enforced.
Whilst changes to the UK’s copyright regime are not required in order for our rights to be enforceable in relation to the training of AI models, it would accelerate the development of the market for AI training content if the government were to issue clarification that the use of copyright materials in the training, fine-tuning and / or grounding of AI systems requires a licence. The Alliance’s preferred policy approach is therefore option 1.
Proposed approach
The introduction of an all-purpose copyright exception for text and data-mining - albeit one with an optout - would detrimentally impact all content businesses that use the internet to distribute and monetise their content.
The government’s proposals discard the principle - long established in copyright law and in-line with natural justice - that the onus to secure legal access for the commercial use of creative works sits with the copying business; it should not be incumbent on the rights-holder to take measures to prevent this. Aside from the in principle issues at play here, there are practical implications which would undermine the ability of creators to control and monetise their works even in the event that they did opt-out via a functioning rights reservation mechanism, as proposed under option 3.
Whilst, under a working opt-out, rights-holders would be able to reserve the rights of the content which they control, they are unable to exercise these rights for ‘downstream’ copies of their IP. Given the proliferation of such content on the internet (for example, screenshots of publisher websites pasted onto social media), this undermines the opt-out and the operation of the policy, more broadly.
Furthermore and for a number of reasons, the negative effects of option 3 on rights holders would fall hardest on individual creators and smaller businesses, such as Alliance members. Firstly, whilst the government’s summary assessment addresses this in outline - and we welcome this - it is hard to conceive of a way in which the technical costs of implementing any opt-out mechanism would not disproportionately fall on smaller content businesses, given the time and costs are likely to be common regardless of enterprise size. Many of our members are already operating with diminishing resources thanks to the devastating impact of the imbalanced relationship between big tech and content creators. This would add to this challenge.
The government’s preferred route would disproportionately impact publishers that predominantly monetise through advertising, as is the case with the majority of Alliance members. Under the government’s proposals, absent exercising their opt-out, subscription publishers would at least be able to command a fee when their content is accessed for AI training. This would not be the case for openaccess publishers. Furthermore, given the absence of paywall barriers, it is likely that there will be a greater volume of downstream works taken from advertising-based publishers versus larger, closedmodel content creators.
Finally, we believe it is important to note that the smallest, and least-established, publishers - such as individual creators - have the least market power and need all means of distribution possible. It is therefore possible that such outlets will be the only ones that do not exercise their-opt out. This would demonetise their content and, taking a wider view, throttle the growth of tomorrow’s content businesses.
Technical standards
The Alliance agrees with the government in welcoming the work that has been done to develop an effective rights reservation mechanism. Unfortunately this has not yet resulted in a technology or standard which performs the necessary functions and has received widespread adoption. We believe the government could have an important role in accelerating this process.
We also agree with the government’s contention that the robots exclusion protocol (‘robots.txt’) is inadequate in serving as a rights reservation mechanism and a replacement must be developed irrespective of the policy direction that follows the conclusion of this consultation.
The development of an effective and operable rights reservation system is absolutely essential if the government does ultimately - and in spite of the damage this would cause to the creative industries - decide to proceed with option 3. The introduction of a broader TDM exception could take place only once such a mechanism has been developed, tested and has proven to function as intended. We note that the press release accompanying the consultation states ‘before these measures could come into effect, further work with both sectors would be needed to ensure any standards and requirements for rights reservation and transparency are effective, accessible, and widely adopted.’ Whilst we welcome this, a further, unambiguous and official policy statement to this effect when the government publishes its response to the consultation would provide much-needed reassurance.
On the specifics of the mechanism itself, we believe that any such rights reservation mechanism must have the following characteristics:
• Legally-enforceable - signals must have unambiguous standing in UK copyright law
• Granular - rights-holders must be able to exercise their rights on a detailed basis in relation to the use of their content, including its storage.
• Practical - reserving rights cannot be an overly time-consuming, expensive or complex process for IP holders. For developers, similarly, signals must be machine-readable and clear.
• Multimedia - the mechanism must be applicable across online media formats (i.e. text, imagery, video, audio).
Transparency, contracts and licensing
The consultation asks respondents to consider what measures should be introduced to support good licensing practices. The primary barrier to a vibrant, liquid and fair licensing market developing is the absence of transparency in the crawling of online content by AI developers and the ongoing violations of copyright laws by AI businesses. It ought to be obvious that a market cannot be expected to develop when theft is a viable alternative to payment.
The practices of ignoring publishers signals delivered via robots.txt and using hidden / undeclared or third-party web crawlers to obtain content illegally are widespread, even by businesses with funding in the billions. These clandestine practices suggest that many AI developers are aware that the use of copyright materials for AI systems requires a licence but that winning the competitive race between them has been deemed to be more important than respecting, and remunerating the creators of the content that is used to build and power these technologies.
Transparency around web-crawling is therefore essential. The Alliance believes that the operators of systems which access and download data from websites on an automated basis should be placed under the following obligations:
• To mandate the use of a consistently-named user-agent string
• To declare the legal entity owning a crawler
• To declare the onward use of data scraped from the internet including the first-party models these data were used to train or fine-tune and the applications data feeds in real-time (if third-party applications are served then the legal entities owning these systems)
In addition to these transparency requirements which apply to the means by which AI developers access content from the internet, developers of foundation models should be required to publish details of all non-proprietary training materials used in model development. This would include individual works (i.e. URLs of specific pages), means of access, date of access and legal basis for access.
Other comments
The consultation makes no mention of historic training and the economic harm that has already been incurred as a result of infringement. Consumers are already adopting AI tools and this is impacting publisher traffic and monetisation. Some publisher content delivers value for the right holder over a long period of time. Where this is the case, the economic harm caused by the theft of this IP and its use in the training corpuses of existing models is likely to be very substantial and persist long into the future. The government must address this harm through its policy development work in this space.
The consultation is focused on model training, however much of the use of IP by AI systems occurs in real-time via grounding or retrieval augmented generation (RAG) processes. These uses of content are equally important, both for the value they deliver for AI developers and the risks presented to rightsholders. The scope of policy development must be expanded to incorporate these data uses.
To conclude and summarise the response of the Alliance, we believe that the UK would be best served by policy aimed at encouraging a licensing market to develop for the use of content by AI systems. This an achievable goal - there is very substantial demand for quality content from AI developers and a willingness from rights-holders to provide access, providing payment is fair and representative of the risks such arrangements entail. The over-100 AI licensing deals that have already been concluded are evidence to this. Such an objective, we believe, requires a clarification of IP law, transparency in the use of content by AI developers and the development of a functioning rights reservation mechanism. We hope the government reconsiders its proposed approach with this overall objective in mind. Independent publishers are content businesses. Our members are not seeking government subsidy or support. Instead all we are hoping to achieve is fair payment for the use of our core product. Once again, the Alliance thanks the government for the opportunity to respond to this consultation and is hopeful that policy moves in a more equitable direction. David Buttle Board Member Independent Publishers Alliance
Comments