Government AI-copyright consultation: only 3% back opt-out exception
The government has released an interim progress report on AI and copyright, alongside a summary of consultation responses required under the Data Act. The consultation closed on 25 February and drew more than 11,500 responses across creators, rights holders, AI developers, researchers, cultural institutions, legal professionals and individuals.
Respondents assessed four options: do nothing; require licensing in all cases; create a broad text and data mining (TDM) exception; or adopt a TDM exception with a rights-holder opt-out plus transparency measures (the government's initial preferred option).
What respondents said
- 88% backed licensing in all cases where data is used for AI training.
- 3% supported the opt-out TDM exception (government's preferred option).
- 0.5% supported a broad TDM exception with no opt-out.
- 7% supported doing nothing and keeping current law.
In short, a decisive majority favours explicit licensing over exceptions. Support for any form of copyright exception totals 3.5%.
Industry reaction
Publishers Association CEO Dan Conway called for the opt-out model to be dropped: "It's clear from the response to the consultation, as released today, that the government must rule out a copyright exception immediately and back the UK's licensing regime with the introduction of transparency requirements. Only 3.5% of respondents backed a copyright exception while the great majority - 88% - want to see more support for licensing."
Authors' Licensing and Collecting Society deputy chief executive Richard Combes said: "These results clearly demonstrate the lack of support for a copyright exception and the pressing need to develop fair and transparent licensing solutions which recognise and respect the rights of authors and other creators."
WGGB general secretary Ellie Peers added: "It's loud and clear, there is widespread opposition from the Government's initial preferred option of an 'opt out' and a copyright exception for text and data mining... UK copyright needs to be enforced not weakened."
Why this matters for government teams
- Policy direction: The centre of gravity is firmly with licensing plus transparency. Prepare for a framework that reinforces permissions, provenance and record-keeping over datasets used for AI training.
- Procurement: Build licensing checks into due diligence for AI systems and datasets. Require vendors to attest to lawful training data and provide clear audit trails.
- Data governance: Maintain inventories of third-party content and licences. Plan for disclosure obligations around sources, dataset composition and opt-out handling.
- Risk management: Assess legal exposure where models or tools rely on scraped or uncertain data sources. Include indemnities and remediation paths in contracts.
- Stakeholder engagement: Expect sustained pressure from creators and collecting societies. Keep lines open with rights holders to pilot workable licensing solutions.
- Budgeting: If licensing prevails, anticipate ongoing costs for access to high-value datasets and tools trained on them.
Timeline and next steps
This is an interim update. The government will publish its report on the use of copyright works in AI development, with an economic impact assessment of the options, by 18 March 2026. Departments should use this window to review procurement, data governance, and contract standards.
For background and policy materials, see the UK IPO's collection on AI and IP policy and research: GOV.UK - AI and IP. For the cross-government approach, see the AI regulation white paper: GOV.UK - AI regulation.
Practical checklist for the next quarter
- Update AI procurement templates to require licensing attestations, dataset summaries and opt-out compliance.
- Map current AI tools and data sources; flag any with unclear provenance for legal review.
- Set up a lightweight transparency register covering models, training data categories and licensing status.
- Engage with collecting societies to explore standard licences and reporting formats.
- Brief leadership on likely budget impact of a licensing-first regime.
If your team needs to build baseline AI literacy to meet upcoming transparency and procurement demands, explore role-based learning paths here: Complete AI Training - Courses by job.
Your membership also unlocks:
