UK AI ambitions are built on a copyright fault line
Britain wants AI sovereignty and resilience, but years of indecision over copyright have left shaky foundations
The legality of AI model training using publicly available copyrighted materials continues to linger like the aftermath of a Labrador’s Christmas day treats: slightly embarrassing, toxic and not going anywhere fast.
Whilst UK thinking around AI is demonstrably shifting from an immature sovereignty debate to a sophisticated and structured conversation on understood risk, resilience and frontier models, much of this improved conversation becomes entirely redundant. At least, it does until the government sorts out its copyright conundrum.
Can models legally be trained in the UK without infringing copyright? We do not know.
The problem is legal certainty. UK policymakers have yet to establish a clear framework governing the use of copyrighted material to train AI models, including how or if existing text and data mining exceptions should apply. That uncertainty affects model developers, rights holders and companies deciding whether to invest in UK AI infrastructure.
Neverending consultation
Three successive governments have failed to make a decision on this issue. Initially, it was sent to the Intellectual Property Office (IPO) in 2022. In February 2024, after two years of consultation, the IPO passed the question back to the now-defunct Department of Science Innovation and Technology (DSIT), saying that the question was too difficult.
This shifted responsibility to the Minister, making it DSIT’s turn to consult - which it did in December 2024, with responses required by February 2025. They received many thousands.
Former Secretary of State Liz Kendall held roundtables for around 60 organisations in September 2025, one of which I joined. She committed to a full report by March and the setup of specialist task forces.
Little to nothing appears to have happened. Then on 18th March 2026, Kendall made a statement to Parliament, backtracking and saying “it will take time to get this right.”
This failure to make a decision quickly around AI and copyright has had consequences. OpenAI backed out of promised funding of the UK Stargate project with NScale back in March. Despite the UK’s light touch regulatory regimen for AI, OpenAI referred to the UK’s failure to act with regard to copyright and model training.
A ministerial letter published 11th November, but dated 30th October, by Ian Murray MP, responded to a public letter from Caroline Dinenage MP. Murray outlines where responsibilities for AI and copyright lie across different departments following DSIT being disbanded. Murray’s repetition of the statement that Parliament would “take its time to get it right” certainly cannot be argued with.
Indeed, it has taken several years and three governments so far. But sometimes inertia caused in the pursuit of perfection is more problematic, and that is where we are today. Murray has promised a further task force will be set up before Christmas.
This is not the first time we have been told it will all be over by Christmas.
Re-framing the question
The UK has a world-leading creative sector and everyone loves it - particularly HMRC. In 2024, the creative, media and sport sectors contributed an estimated £247.6 billion to the UK economy, according to DCMS. Conversely, the AI sector’s revenues increased by 41% to £33.6 billion in 2025 and employed an estimated 130,400 people.
MPs have an impossible task, deciding between their existing favoured child and the potential AI future. Both are, of course, very important. This was exactly the basis of the House of Lords report in March this year. Trying to please everyone has pleased nobody. Without a mind-shift in approach this will simply never reach resolution, and nothing in the recent activity has given any indication of a change any time soon.
But the conversation ought never to have been framed as an argument between creatives and tech. This framing is fundamental to the failure to make a decision.
The government that fixes this sets its nation up for success
The issue isn’t creative versus tech. A great deal of innovation, whether through research content or open source software licensing, is also used to train AI. AI is core to innovation in many industries and the creative industries are widely adopting and benefitting from it, like the rest of the UK. Only last week the Office of National Statistics shared growth of 0.7% in GDP attributed to AI and cloud computing.
Instead of creatives versus tech, the conversation could be a question of how we balance existing legal regimes to allow AI innovation whilst ensuring that we protect human content creators – whether scientific, innovative or creative.
For open source the debate is a difficult line to tread in. Our industry is based on accreditation and respect for contribution, but it has also evolved out of copyright restrictions over the years.
In reality we need new business models to protect human content creators of the future. I have argued many times that the copyright debate becomes a distraction. Rather than looking at how to create new ways to run businesses around creative industries, holding on to copyright alone risks losing out to other businesses that solve the problem from a different market.
The prime example of this is streaming, where a technology company (Spotify) has taken the lion’s share of the market, leaving those that create music with a smaller share than they had previously.
Turing report fails at the copyright hurdle
As part of its new report, Securing resilience for the age of AI, The Turing Institute focused on five risks in frontier models: cybersecurity risks; democratic instability; misalignment; loss of meaningful human control; and chemical and biological threats.
The report suggests the use of a standard in respect of the copyright confusion. That’s laudable, and well-made open standards are great. But they are not made in lieu of law, but to implement it.
A report launch event panel questioned on this saw a representative of Sov.AI fluff the issue, saying that they comply with laws. Others panelists were able to do nothing more than point to the report. Lucy Yu’s comment that “it’s later than we think” rang all too true.
The report’s approach to resilience is heading in the right direction and focused on some of the key areas of risk, acknowledging the overall supply chain. But our thinking must be more holistic and understand that the lack of a robust solution to this key challenge undermines the landscape. The last thing we need is more consultation.
Meanwhile the IP Office is offering £80k in a tender for a survey of the creative industries. It might be seen as a small price to pay to make the never-ending problem go away - but it won’t solve the problem, just kick the issue even further down the road. Only a clear decision on the future around copyright can do that.
Getty v Stability AI could have global implications
Decisions in the UK have the potential for international impact too. The IP Federation has sought to comment on the Getty Images v Stability AI appeal: “At the heart of the appeal is the question of whether an AI model developed overseas can be treated as an infringing copy under UK copyright law”.
The IPF suggests the courts should consider the wider interests of British industry saying that the case’s outcome “could have far reaching implications for British industry and the UK's ambitions to become a global leader in AI." The case will decide “whether an AI model developed overseas can be treated as an infringing copy under UK copyright law, even where the model itself does not contain or reproduce the copyright works in question."
Whilst the UK is not alone in struggling with the balance between AI innovation and copyright issues, other countries have created some level of certainty on an interim basis at least. This has made it easier for companies to plan ahead.
Industry has given government time and input, and it really is time that the government gets on with it and gives us clarity.
Amanda Brock is CEO of OpenUK.