Out-Law News 5 min. read
31 Aug 2023, 3:51 pm
A new code of practice aimed at balancing the rights of content creators with the desire of AI developers to train their systems using quality data is set to be finalised this autumn, Out-Law understands.
The code is currently being developed by a working group established by the UK’s Intellectual Property Office (IPO), which consists of representatives from the technology, creative and research sectors. The working group met for the first time in June.
The terms of reference of the working group are published on the IPO’s website. The IPO has said the code will aim to “make licences for data mining more available”, helping “to overcome barriers that AI firms and users currently face, and ensure there are protections for rights holders”. It expects “parties to enter into the final code of practice on a voluntary basis” but it has said legislative reform could be considered if the code is not agreed or adopted.
Progress towards a new code of practice comes as a prominent group of MPs called on the UK government not to expand the existing text and data mining exception in UK copyright law in pursuit of its aims of supporting the UK AI industry to innovate and grow. In a report published on Wednesday, the Culture, Media and Sport Committee said the government “must work to regain the trust of the creative industries following its abortive attempt to introduce a broad text and data mining exemption”.
Cerys Wyn Davies
Generative AI challenges the traditional view of text and data mining as a technique that can be used for the purpose of learning – it shows how information that is mined can also be essentially regenerated via fresh outputs produced by generative AI systems
Cerys Wyn Davies of Pinsent Masons, who specialises in the application of intellectual property law to new technologies like AI, said: “The debate over how to appropriately support AI innovation while protecting the interests of rights holders has taken on greater significance in the last year with the emergence of increasingly popular generative AI systems like ChatGPT.”
“Generative AI challenges the traditional view of text and data mining as a technique that can be used for the purpose of learning – it shows how information that is mined can also be essentially regenerated via fresh outputs produced by generative AI systems. Rights holders are concerned about the risks this poses to their business models, but it also raises fundamental questions relevant to AI development too,” she said.
“AI relies on quality data sets to produce safe, useful and reliable outputs. If value cannot be maintained in content creation, there will be fewer incentives for content creators to make free-to-the-public the content they produce, with the knock-on risk that AI systems rely on inaccurate or otherwise skewed data – and that this leads to inherent bias in the way those systems operate and bad outcomes for users. It is in this high-stakes environment that work on the new code of practice is taking place, with the legal action taken by Getty Images against Stability AI also indicative of action rights holders might take against AI developers to enforce their rights,” she said.
The Copyright, Designs and Patents Act 1998 currently provides limited exceptions to copyright infringement. One exception allows text and data mining (TDM) of copyrighted works for non-commercial purposes, provided that the user has lawful access to the work via, for example, a licence, subscription or permission in terms and conditions. Last summer, following an earlier consultation, the UK government set out plans to expand the scope of that exception to enable TDM of works protected by copyright and database rights for any purpose after considering that it could support AI and wider innovation in the UK.
Those proposals drew criticism from rights holder groups, however. After considering the feedback from industry, the government revised its policy position, with different government ministers indicating that the planned reforms would not be taken forward. A spokesperson for the IPO has now confirmed to Out-Law that legislative reform is not being pursued – and that its efforts are currently focused on facilitating the industry-led voluntary code on copyright and AI instead.
“As was announced last year, the government will not proceed with its previous policy for a broad copyright exception for data mining,” the IPO spokesperson said. “Instead, it seeks to strike a balanced and pragmatic approach through the development of a code of practice which allows AI innovators and the creative industries to grow in partnership. This supports the government’s ambition to make the UK a world leader in research and AI innovation, whilst ensuring our copyright framework continues to promote and reward investment in the UK’s world-class creative industries.”
The spokesperson said the working group “continues to make progress in its discussions” and that the IPO believes the involvement of representatives from both the AI and creative sectors “will lead to an agreement that identifies, develops, and codifies good practice on the use of copyright, performance, and database material in relation to AI – including data mining”.
Senior Practice Development Lawyer
It is not yet clear what practical solutions will emerge from the output of [the working group’s] work
Work on the development of the new code comes after the government confirmed in March that it endorsed recommendations made by its former chief scientific adviser – including that it work with the AI and creative industries to “develop ways to enable TDM for any purpose, and to include the use of publicly available content including that covered by intellectual property as an input to TDM (including databases)”. Sir Patrick Vallance’s recommendations followed a review he undertook into pro-innovation regulation of digital technologies.
Copyright law expert Gill Dennis of Pinsent Masons said: “The terms of reference of the working group developing the code that have been published are vague in the sense that it is not yet clear what practical solutions will emerge from the output of their work. AI developers will, for example, be keen to avoid any heavy burden that would come from having to agree individual licensing agreements with each and every content creator that its AI system could scrape information belonging to, while rights holders in turn will want to ensure they are fairly compensated for how their works are used – and that there is sufficient transparency over, and monitoring of, such use.”
Dennis said some stakeholders may be sceptical as to whether a non-legislative initiative such as an industry-led voluntary code of practice can strike the right balance of interests and resolve the practical and financial conundrums at issue. She cited another report published by a committee of MPs this week in support of that notion.
In that report, the Science, Innovation and Technology Committee called on the government to accelerate the creation of a new AI governance regime to address 12 challenges AI poses – including in relation to intellectual property. The MPs cited criticism from The Library and Archives Copyright Alliance of the government’s decision not to take forward plans for a broad TDM copyright exception, which said the decision “prevents the UK from capitalising on the diverse, agile and creative benefits that AI can bring to the UK’s economy, its society and its competitive research environment”.
The Science, Innovation and Technology Committee further warned that the UK risks being left behind as a standard-setter in relation to AI if policy is developed in other jurisdictions first. It urged the government to introduce ‘whatever statutory measures as may be needed’ as part of that new framework. The government published its AI white paper earlier this year in which it outlined plans to retain the current sector-based approach to the regulation of AI in the UK but with an overlay of cross-sector non-statutory principles that regulators would have to “interpret and apply to AI within their remits”.
The UK government is seeking to play a leading role in the development of a more harmonised framework governing use of AI globally. Prime minister Rishi Sunak is due to host an AI safety summit in the UK in November.
In a letter (2-page / 67KB PDF) published on Thursday, Dan Conway, chief executive of the Publishers Association – a body that represents more than 140 UK publishers – urged Sunak to “make clear that UK intellectual property law should be respected when any content is ingested by AI systems and a licence obtained in advance”, either as part of, or in parallel with, the November summit.
Conway, who previously described the now-rejected plans to broaden the TDM copyright exception as “a sledgehammer to crack a nut”, said in his letter: “The training of AI systems should be done transparently, with the consent of, and in a manner that credits and fairly compensates the creator or IP rightsholder i.e. under licence.”
26 Jan 2023
01 Mar 2023