UK Government's report on copyright and AI

The question

How does the UK Government’s latest AI and Copyright report and impact assessment shape the legal and policy environment for generative AI and large platforms, and what should major tech businesses do now?

The key takeaway

The Government has reconsidered its original preference for a broad text and data mining (TDM) copyright exception and instead will focus on enforcing the existing UK copyright framework. This will focus on developing transparency, technical standards and “best practices”, and exploring new personality‑style protections for digital replicas. Large platforms should assume UK‑facing AI products will need robust copyright compliance, training‑data transparency, technical controls for web crawling, and stronger governance around deepfakes and replica uses.

The background

Following strong lobbying from both creative industries and AI developers, the UK consulted on four main options for copyright and AI policy: maintaining the status quo; strengthening copyright - requiring licensing in all cases; a broad commercial TDM exception and an EU‑style data‑mining exception with a rights‑holder opt‑out option.

81% of respondents to the consultation preferred “stronger copyright”, while the broad exception with an opt‑out had the lowest level of support (3% of respondents). This was a significant finding as the TDM exception with an opt-out exception was the Government's original preference. 

The development

No new broad TDM exception (for now)
The Government has explicitly moved away from its previous preferred option of an EU‑style TDM exception with rights‑holder opt‑out, citing strong creative‑sector opposition, gaps in evidence on economic impacts and technical feasibility, and uncertainties about how copyright impacts the wider development and deployment of AI. Instead, the Government will leave core copyright provisions unchanged for now, while exploring more “focused” exceptions, such as for research or specific public‑interest uses such as safety and security once further consideration has been given to the effects of proposals on copyright owners, developers, and users.

Considering input and output transparency and technical standards
 The report establishes a clear policy direction rather than an immediate suggestion of following the route of the EU by adopting a transparency regime. Over 90% of respondents agreed that AI developers should disclose the sources of their training material. Creative industry respondents and technology companies similarly regard greater transparency as key to the effective licensing of copyright material and enforcement of IP rights. Despite alignment on the importance of transparency, views vary in relation to potential disclosure requirements. The consultation highlights differing views between the creative industry and technology industry with tech firms being the ones who generally supported higher-level, voluntary disclosure. The Government intends to continue monitoring the effects of transparency rules in other countries and to keep under review what changes, if any, may be appropriate to introduce in the UK.

Computer‑generated works (CGWs): likely removal of unique UK right
In the UK, copyright protection is available for artistic work which is "generated by a computer in circumstances such that there is no human author" through section 9(3) of the Copyright, Designs and Patents Act 1988 (CDPA). The "author" of such a work is usually the person "by whom the arrangements necessary for the creation of the work are undertaken". In the case of generative AI, the "author" will usually be the person who inputs the prompt(s). The Government's report notes that 78% of online respondents were not in favour of maintaining the current CGW protection, agreeing that there should not be protection for works created solely by AI.  The common reasoning being that the maintenance of the current CGW creates competition between CGWs and human works whereas removing copyright in CGWs would eliminate this competition. The Government’s preferred option remains to repeal section 9(3) CDPA, leaving AI‑assisted works protected where they reflect human intellectual creation. This position aligns with the Government's sentiment that copyright should incentivise and protect human creativity, while balancing interests in innovation and technological development.

Digital replicas and possible new personality‑style right
The Government recognises that current UK IP law gives limited control over a person’s image, voice or “style”, and that current intellectual property rights only partly address non‑consensual digital replicas. Given evidence of significant harms (such as, commercial imitation, reputational damage, and tech‑facilitated sexual abuse), the Government will conduct further consultations to explore new protections. This will potentially include a standalone “personality right” or a “digital replica” right to more sufficiently address unauthorised digital replicas. With explicit recognition that any reform must be proportionate and avoid unintended constraints, the government will monitor international developments (such as the proposed ELVIS Act and NO FAKES Act in the US, and emerging EU approaches) to help identify the potential impact of supplementing the current copyright legal framework.

Why is this important?

For large platforms developing (and/or deploying) foundation models relying on large‑scale use of web‑scraped material for training, the headline is that the UK will not offer a “copyright‑light” haven for AI development. Developers should assume that:

• UK law requires licences (or a valid exception) for copies made during training and fine‑tuning;
• any new digital‑replica or personality right is likely to impact product, content moderation, and advertising considerations for platforms hosting or generating synthetic media;
• imported models trained abroad may still face UK secondary‑infringement risk if they are treated as “articles” containing infringing copies; and
• transparency and adherence to technical standards will be central to UK and EU expectations, even if not yet fully codified in statute.

Any practical tips?

For now, major platform clients should treat the report as a clear signal of regulatory expectations and consider aligning internal frameworks accordingly. Practical steps to consider taking include:

• mapping UK‑relevant training: document which datasets (including user‑generated content, scraped web data and licensed collections) are used to train or fine‑tune models accessible in the UK, and where that training occurs;
• tightening crawler governance: ensure all AI/web crawlers are clearly identified and log access for audit and potential rights‑holder challenges;
• build a training‑data transparency position: even in the absence of UK‑specific rules, establish robust systems and processes, for example, ensuring there is an adequate process to respond to rights‑holder queries;
• review licensing and opt‑out strategy: continue negotiating licences with key sectors (music, images, news, books etc.) and track emerging collective licensing schemes; design processes to honour machine‑readable “do not train” signals where feasible.
• prepare for digital‑replica regulation: audit products enabling voice/face cloning “in the style of” prompts or virtual influencers; and enhance consent, labelling and takedown processes. 

Summer 2026