UK Government Abandons AI Training Copyright Exception After 88 Percent of Consultation Respondents Backed Mandatory Licensing
The UK government published its statutory report on copyright and AI on March 18, dropping its preferred opt-out exception for AI training after overwhelming opposition and signaling a shift toward market-led licensing and transparency obligations.
The UK government on March 18 published a 125-page report and economic impact assessment on the use of copyright works in AI development, fulfilling a statutory deadline set by the Data (Use and Access) Act 2025 and confirming a significant policy reversal: it will no longer pursue a broad copyright exception that would have allowed AI developers to train on protected works unless rightsholders explicitly opted out.
The report, produced jointly by the Department for Science, Innovation and Technology, the Department for Culture, Media and Sport, and the Intellectual Property Office, marks the conclusion of one of the most heavily responded-to government consultations in recent years. More than 11,500 submissions were received between December 2024 and February 2025, with 88 percent of respondents supporting a requirement for licensing in all cases and just 3 percent backing the government’s original preferred option of a text and data mining exception with rights reservation.
Four Options, One Clear Rejection
The consultation had presented four policy paths. Option 0 would have maintained existing copyright law unchanged. Option 1 required licensing for all AI training uses of copyrighted material. Option 2 proposed a broad data mining exception with no rights reservation. Option 3 — the government’s stated preference — would have created a text and data mining exception while allowing rightsholders to reserve their rights through an opt-out mechanism.
The sectoral divide was stark. Creative industries, which contribute £124 billion to the UK economy annually and employ 2.4 million people, overwhelmingly favored mandatory licensing. Technology companies leaned toward the exception-based approaches. But the scale of opposition to Option 3 — encompassing publishers, musicians, visual artists, trade unions, and academic institutions — forced the government to abandon it as its preferred way forward.
Licensing Market Takes Priority
Rather than legislating immediately, the government is placing its confidence in a market-led approach to licensing. A growing number of commercial deals between AI firms and content owners — particularly in publishing, music, and image libraries — suggests that a licensing ecosystem is beginning to emerge, though the report acknowledges the market remains “new and evolving” and lacks transparency.
The government stated it will not introduce copyright law changes until confident that reforms meet both economic and public interest objectives. It is monitoring international frameworks, including the EU AI Act’s transparency provisions, and has committed to considering future legislation if the voluntary approach proves insufficient.
Four expert technical working groups, comprising more than 50 specialists, continue to examine practical solutions across control and technical standards, information and transparency, licensing mechanisms, and wider support measures for the creative sector.
Transparency and New Protections
More than 90 percent of consultation respondents supported requirements for AI developers to disclose the sources of their training data. The government proposes developing industry-led best practice on input transparency as an initial step, with statutory measures held in reserve. The report also explores introducing personality rights to address the risk of realistic AI-enabled impersonation — a new area of protection that the UK’s existing intellectual property framework does not cover.
House of Lords Pushes for Stronger Action
The government report arrives less than two weeks after the House of Lords Communications and Digital Committee published its own inquiry, “AI, Copyright and the Creative Industries,” on March 6. The Lords committee adopted a harder line, recommending that the government explicitly rule out any commercial text and data mining exception and instead adopt a “licensing-first” approach backed by statutory transparency obligations and a dedicated regulatory body to enforce compliance.
The committee described the situation facing creators as “a clear and present danger from uncredited and unremunerated use of copyrighted material to train AI models,” noting that photographers, musicians, and authors are seeing AI systems trained on their work produce imitations that directly compete for their employment and earnings. It called on the government to finalize its approach within 12 months and to prioritize the development of sovereign AI models as an alternative to reliance on opaquely trained systems developed in the United States.
A Holding Position
The March 18 report sets a direction but does not resolve the underlying tension between creative industry protection and AI development. The government has effectively adopted a wait-and-see stance, betting that commercial licensing can scale quickly enough to compensate rightsholders fairly while the technology sector retains access to training data. If that market fails to materialize on equitable terms, legislative intervention remains on the table — but no timeline has been set for that determination.