AI and copyright: what businesses need to know

The current UK legal position on AI and copyright, including training data risks, AI-generated content ownership, and pending government reforms.

UK-wide
Guide summary

Check if your AI training data has copyright permission. You may own AI-generated content under current law, but this is uncertain. Watch for new rules coming in 2026.

  • Check AI provider terms for training data copyright compliance
  • Get legal advice before using copyrighted material to train AI
  • Commercial AI training needs copyright holder permission
  • AI-generated content may be owned by the person who arranged its creation
  • Copyright for AI works lasts 50 years from creation
  • Government report on AI and copyright due by 18 March 2026
  • New AI copyright laws expected in late 2026
  • Current UK law does not cover modern AI systems clearly
On this page
UK-wide

Related guides in Compliance & Legal

Software licensing compliance

Understand your legal obligations when using, developing, or distributing software - including open source licensing, commercial agreements, and …

AI Regulation Framework

The UK takes a principles-based, sector-specific approach to AI regulation. There is no single AI law. Instead, existing …

Artificial intelligence raises significant copyright questions for businesses. There is no AI-specific copyright law in the UK yet. The government is actively developing policy through consultation, industry working groups, and a statutory reporting obligation under the Data (Use and Access) Act 2025.

The key areas businesses need to understand are:

  • training data rights — can you legally use copyrighted material to train AI models?
  • AI-generated content ownership — who owns the output from generative AI tools?
  • pending reforms — what changes are coming and when?

This guide explains the current position and what your business should do now.

Current legal position

The lack of AI-specific copyright legislation creates uncertainty for businesses. Existing copyright law under the Copyright, Designs and Patents Act 1988 (CDPA) was not designed with generative AI in mind. Courts have not yet tested how these rules apply to modern AI systems.

The government ran a consultation on AI and copyright from December 2024 to February 2025. The consultation attracted significant interest from both the creative industries and the technology sector. In December 2025, the government published a statement of progress confirming that two industry working groups had been established to develop practical solutions around transparency (how AI developers disclose what training data they use) and technical measures (how rights holders can control use of their content).

The DUAA 2025 (sections 135–137) requires the Secretary of State to produce a copyright report and economic impact assessment by 18 March 2026. This report will shape future legislation.

Training data risks

The practical implications for businesses are significant. If you use third-party AI tools such as large language models, image generators, or code assistants, you should check the provider’s terms of service regarding copyright compliance of their training data. Some providers have secured licences from rights holders. Others have not, and this creates downstream risk for your business.

If your business is training its own AI models, seek legal advice before using copyrighted material. The UK’s text and data mining exception under CDPA 1988 section 29A applies only to non-commercial research. There is no broad commercial exception. This is different from the position in some other jurisdictions.

Key risk areas include:

  • Scraping web content to build training datasets without permission
  • Fine-tuning models on proprietary or licensed content
  • Using AI outputs that closely reproduce copyrighted source material

AI-generated content ownership

The CDPA 1988 section 9(3) provides for copyright in computer-generated works — works generated by a computer in circumstances where there is no human author. Copyright belongs to the person who made the arrangements necessary for the creation of the work.

This provision is unusual. Most jurisdictions require human authorship for copyright to exist. The United States, for example, has ruled that purely AI-generated works cannot be copyrighted.

For computer-generated works, copyright protection lasts 50 years from the date the work was made. This is shorter than the standard life-plus-70-years term for works with a human author.

However, the application of section 9(3) to modern generative AI is untested in court. If your business creates content using AI tools, you may own copyright under this provision, but there is no guarantee. Key uncertainties include:

  • Who counts as the person who “made the arrangements” — the user, the AI developer, or both?
  • Whether prompting an AI model constitutes sufficient “arrangements” for copyright to arise
  • Whether works with significant human editing or curation have a human author (and therefore get the longer protection term)

Until courts or legislation clarify these questions, treat AI-generated content with caution. Do not rely on copyright protection for commercially valuable AI-generated works without legal advice.

What’s changing

Several developments will shape the future of AI and copyright in the UK:

  • Copyright report (March 2026): DUAA sections 135–137 require the Secretary of State to report on AI and copyright, including an economic impact assessment. This is due by 18 March 2026 and will inform future legislation.
  • Government AI Bill (second half of 2026): Expected to include provisions on AI and copyright. The scope and content will depend on the copyright report and working group outcomes.
  • Industry working groups: Two groups are developing practical solutions — one on transparency (how AI developers disclose training data use) and one on technical measures (how rights holders can control access to their content).
  • EU position: The EU has broader text and data mining exceptions under the Digital Single Market Directive (Articles 3–4). Rights holders can opt out of commercial TDM. UK businesses operating across both markets need to understand both regimes.

The direction of travel suggests some form of transparency obligation for AI developers is likely. Whether the UK introduces a broader commercial TDM exception or a licensing framework remains to be seen.

Practical steps for businesses

  1. 1

    Audit your AI tools

    Review the terms of service for any AI tools your business uses. Check whether the provider has secured licences for training data and what indemnities they offer against copyright claims.

  2. 2

    Document AI-generated content

    Keep records of how content was created, what tools and prompts were used, and what human editing was applied. This supports any future copyright ownership claims and demonstrates good practice.

  3. 3

    Do not assume you can train AI on copyrighted material

    The UK has no broad commercial text and data mining exception. If your business trains its own models, seek legal advice on licensing before using copyrighted content.

  4. 4

    Monitor the copyright report

    The DUAA copyright report is due by 18 March 2026. It will shape future rules on training data, transparency, and licensing. Review its recommendations when published.

  5. 5

    Seek legal advice for high-value uses

    If your business relies on AI-generated content for commercial purposes or is training its own models, get specialist legal advice. The law is unsettled and the risks can be significant.