On 18 March 2026, the UK government published its long-awaited report on UK AI copyright — and resolved almost nothing. DSIT, DCMS and the Intellectual Property Office formally abandoned their previously preferred policy direction, committed to no new legislation, and adopted an explicit wait-and-see position. For businesses deploying or procuring AI tools, the uncertainty this creates is operational, not theoretical.

UK AI Copyright 2026: Signal Summary

On 18 March 2026, DSIT, DCMS and the Intellectual Property Office jointly published the Report on Copyright and Artificial Intelligence, fulfilling a statutory obligation under Sections 135 and 136 of the Data (Use and Access) Act 2025. The report was accompanied by an economic Impact Assessment and a Written Ministerial Statement from Secretary of State Liz Kendall.

The government’s previously preferred approach — a broad text-and-data-mining (TDM) exception with an opt-out mechanism (Option 3) — was formally abandoned. No new legislation was announced. The Copyright, Designs and Patents Act 1988 (CDPA 1988) was not amended. No new regulator was created. The government committed instead to gathering further evidence, monitoring international developments, observing ongoing litigation, and assessing market activity before committing to any specific course of action.

Why This Is Structurally Significant

The rejection of Option 3 matters not just as a policy reversal but as a structural signal. The government had indicated its preference for a TDM opt-out as recently as December 2025. Its reversal, under pressure from 11,500-plus consultation respondents, leaves both AI developers and rights holders without a clear framework — and no timeline for when one will arrive.

The absence of clarity has direct commercial consequences. Leading AI developers have consistently avoided training models in the UK because the legal position is uncertain, choosing instead jurisdictions with more permissive rules. The report acknowledges this explicitly. A wait-and-see position entrenches that dynamic rather than resolving it. For analysis of how UK AI policy uncertainty is repricing technology investment risk, the structural dynamics are examined in full.

The tension is sharpened by timing. The Chancellor’s statement on 17 March 2026 — the day before the report’s publication — committed to the UK achieving the fastest AI adoption in the G7, supported by £2.5 billion in investment. A policy vacuum on the legal basis for AI training sits uncomfortably alongside that ambition.

What Drove This

Three pressures converged to produce a UK AI copyright standoff rather than a decision.

The UK AI copyright consultation received 11,500-plus responses; 88% supported requiring licences in all cases, with only 3% backing Option 3. Creative industries — representing £146 billion in gross value added in 2024, approximately 6% of the UK economy (IPO Impact Assessment, 18 March 2026) — were uniformly opposed to the opt-out approach.

The evidence base was also found to be insufficient for action. The Impact Assessment acknowledged that the economic case for moving decisively in either direction was weak, and that the international context remained volatile — particularly the direction of US and EU approaches.

Then there is the litigation. Getty Images v Stability AI, the first and only AI copyright case in UK courts, remains unresolved. Getty abandoned its primary copyright infringement claims due to insufficient evidence that training or development took place in the UK — demonstrating precisely the enforcement gap that any reform must address. The government cited the case as a reason to wait for judicial clarity before committing.

Implications for UK Businesses

The absence of reform is itself a structural position that demands a response.

For AI developers and technology vendors training models on UK-originated content, the current TDM exception under Section 29A of the CDPA 1988 permits non-commercial research only. Commercial AI training on copyright-protected works remains legally exposed. Proactively documenting training datasets and establishing licensing agreements remains the recommended mitigation — the report created no new permissions and no new protections.

Enterprise buyers procuring AI tools face a more immediate question: vendor liability. AI tools built on models trained without clear UK AI copyright licensing carry exposure that may fall to the buyer through indemnity arrangements — not just the vendor. Procurement due diligence must now include evidence of lawful data sources, audit rights, and contractual protections against infringement claims. This mirrors the governance gap UK businesses face when deploying AI built on unverified training data — a structural accountability problem that predates this report and is not resolved by it.

Creative and media businesses retain their existing rights under the CDPA 1988 — but enforcement is practically difficult without transparency obligations on training data. The Getty v Stability AI case made this concrete: without knowing where training occurred, primary infringement claims are difficult to establish in a UK court.

What to Watch Next

Four forward indicators define the UK AI copyright landscape through the remainder of 2026.

The King’s Speech on 13 May 2026 may include a UK AI Bill. Expected scope is narrow — frontier models and copyright provisions only — and even a bill introduced in May would not become law before late 2026 or early 2027. Its absence from the speech would confirm the status quo extends further.

The Getty Images v Stability AI appeal is expected to produce the UK’s first judicial precedent on AI training and copyright infringement — specifically whether the existing CDPA framework offers any practical protection for rights holders whose work has been used in training on UK-originated material.

The government’s Creative Content Exchange pilot is planned for summer 2026, trialling licensing mechanisms for AI developers to access digitised content from the National Archives, Historic England, and several public institutions. If successful, it represents the first workable model for lawful AI training data in the UK.

Finally, the EU divergence demands attention. Article 53(1)(d) of the EU AI Act requires AI providers to publish a sufficiently detailed summary of training data sources — an obligation with no UK equivalent. UK businesses with EU clients or operations already face a dual-track compliance environment; that gap widens with each month the UK position remains unresolved. For the full picture on how the UK and EU are diverging on AI transparency and compliance obligations, that analysis covers both regulatory tracks.

ObvioTech Assessment

The UK government’s March 2026 report is best understood as an orderly retreat from a position that was untenable — not as a coherent strategy for what comes next. Abandoning Option 3 was correct: a broad TDM opt-out would have imposed verifiable costs on the creative sector while delivering uncertain benefits to AI developers who train outside the UK regardless of domestic law.

The problem is that the absence of any alternative framework is not a neutral outcome. It actively disadvantages UK-based AI companies competing against developers operating under clearer regimes in the US, EU, and Japan — and it was true before this report, which has changed nothing in practice.

For operators, the implication is direct: UK AI copyright uncertainty is a present operational condition, not a future risk to be managed when legislation arrives. Organisations that build copyright governance into procurement frameworks, vendor contracts, and AI deployment policies now will be materially better positioned when the legal landscape finally settles — whenever that is.

Sources

— DSIT / DCMS / Intellectual Property Office. Report on Copyright and Artificial Intelligence. Policy Paper. 18 March 2026.

— DSIT / IPO. Copyright and AI: Economic Impact Assessment. Published under Sections 135–136 of the Data (Use and Access) Act 2025. 18 March 2026. Available via GOV.UK.

— Hansard. Written Ministerial Statement: Copyright and AI Progress. Secretary of State Liz Kendall. 18 March 2026.

— UK Parliament, Communications and Digital Committee. AI, Copyright and the Creative Industries. 6 March 2026.