Today, the Data (Use and Access) Bill will be debated in the House of Commons, during which MPs will likely vote on amendments that would, among other things, mandate transparency requirements for AI developers, allowing rightsholders to discover if their content has been scraped from the internet and used to train a generative AI model.

The amendments, as introduced previously by Baroness Kidron in the House of Lords, is a golden opportunity for the government and MPs, on all sides, to take bold action to improve AI safety, support the UK’s copyright framework and "turbo charge" the licensing of content and future growth in the creative industry.

In my role as CEO at Publishers’ Licensing Services (PLS), a UK not-for-profit representing over 4,500 publishing organisations, I’ve seen at close quarters the frustration, anger – and, particularly for smaller publishers, the deep sense of powerlessness – over the unlicensed use of their content by generative AI models.

Last month, PLS, along with our partners at the Authors’ Licensing and Collecting Society (ALCS) and the Copyright Licensing Agency (CLA), unveiled plans to develop a voluntary collective licence for the use of published content in generative AI models. This builds upon the ever-increasing amount of direct licensing taking place between publishers and AI developers. The announcement was the result of work PLS has carried out over the past year with our partners and through consulting closely with publishers to gain insight and understanding into the opportunities and barriers to licensing content for AI.

In conversations I’ve had with publishers, large and small, one theme came through clearly: without transparency, there can be no trust. While there is willingness to license both directly and collectively, concern exists about the inability to know precisely what content has been ingested, from where it has been obtained, and how it has been used. This asymmetry of information is suppressing the development of a dynamic licensing market for AI, which in turn is hindering opportunity for future growth in both AI and the creative industries.

Large tech companies’ resistance to embracing transparency has only served to further erode and undermine trust between the creative and AI sectors. Indeed, it is only through legal proceedings in the US that we have learned that Meta has trained its large language model using the LibGen database, a dataset of millions of pirated books and academic articles

In their respective responses to the government’s recent consultation on AI and copyright, Open AI argued that disclosing training data would incentivise rightsholders to seek to license content used or look to litigate where a licence has not been obtained (in other words, to legitimately enforce their rights), with Google arguing disingenuously that certain transparency requirements and disclosing training data would potentially risk trade secrets.

It would be unfair to paint all AI developers with the same brush. There are developers who recognise the value of licensed content, the need to renumerate rightsholders for its use, and understand the importance of accountability. It is those tech innovators who the government should be actively championing and assisting. By protecting and building upon the content produced by the UK’s established and incredibly successful creative industries and looking towards the next generation of language models that may not require huge quantities of data and energy, the government can position the UK as a hub for ethical, safe AI.

The call for proper accountability is not anti-tech nor anti-innovation, rather it is a recognition that as generative AI plays an increasingly significant role in our lives, the accuracy and quality of information has never been more important

However, the government’s favoured approach of tying the introduction of any transparency of AI models to the introduction of an "opt out’" copyright exception has meant that building trustworthy and accountable AI has appeared as a secondary consideration.

Included in a package of measures proposed by the government, transparency of AI models has effectively been used as a bargaining chip with the creative industries, to sweeten any damage that the weakening of copyright that their proposed exception would cause, rather than as a fundamental principle of any legislative solution. While I welcome the government’s own amendments committing to publishing an economic impact assessment, this will only lock them into a timeline that will see no transparency and accountability of AI for some time. Any delay to the introduction of transparency measures means that the rapid unchecked development of AI will make doing so increasingly difficult, time-consuming and likely ineffective.

The call for proper accountability is not anti-tech nor anti-innovation, rather it is a recognition that, as generative AI plays an increasingly significant role in our lives, the accuracy and quality of information has never been more important – something that publishers know only too well. It is as unsettling as it is remarkable that a technology, whose adoption is being strongly encouraged by government, remains largely opaque, secretive, and seemingly above the law.

By shining a light on the content used to train a model, transparency can help remove erroneous information and mitigate against bias and harmful AI outputs.

With government hesitancy to introduce any regulation, huge credit should go to Baroness Kidron who recognised the opportunity the Data Bill presented to both highlight the creative industries’ concerns with mass copyright infringement, but also to act to introduce law that would vastly improve AI safety.

It has been reassuring to hear support from politicians from all sides as the bill has progressed through parliament, with transparency amendments introduced by Lib Dem MP Victoria Collins and Labour’s Alex Sobel MP ensuring that this issue is brought to the floor of the House of Commons. It is highly likely that cross-party pressure has been a significant factor in the government’s apparent rethink of its plans.

Speaking on behalf of an organisation that represents publishers eager to license content responsibly and collaboratively, I urge the government and MPs on all sides to seize this opportunity. By voting for the amendments, MPs can ensure that AI developed and used in the UK is safe and accurate and can cement a more sustainable forward-looking relationship between the creative industries and AI sector that capitalises on the willingness of publishers to license high-quality, curated content to drive innovation and unleash potential growth.

It’s a win-win we can all support.