- US: the Copyright Office says purely AI-generated material is not protected and prompts alone do not give enough control.
- US: human creative input, selection, arrangement and edits in AI-assisted work can be protected.
- EU: protection requires the author's own intellectual creation, which points to human creative choices.
- Many business AI contracts assign output rights to the customer; some add an indemnity with conditions.
- Output that reproduces a protected work can infringe, whoever pressed the button.
On this page:United States: human authorshipEuropean Union: the author's own intellectual creationCan AI output infringe someone else's copyright?What vendor terms usually say
United States: human authorship
The US Copyright Office set out its position in Part 2 of its report Copyright and Artificial Intelligence (January 2025). Its main conclusions:
- Using AI to assist, rather than stand in for, human creativity does not affect copyright protection.
- Copyright does not extend to purely AI-generated material, or material where a person had too little control over the expressive elements.
- With today's generally available technology, prompts alone do not give enough control.
- Human contributions that are perceptible in the output, and creative selection, arrangement or modification of AI material, can be protected. This is judged case by case.
When registering a work that contains more than a small amount of AI-generated material, applicants must disclose it and describe the human contribution, under the Office's 2023 registration guidance.
Want to check how your teams use AI-made content before it goes out? Book a free AI Value Briefing and we will look at it with you.
European Union: the author's own intellectual creation
EU copyright protects works that are the author's own intellectual creation, a standard set by the Court of Justice of the EU. A 2025 study for the European Parliament explains that this points to human creative choices, so outputs generated entirely by AI without meaningful human input typically do not qualify, and it notes the legal status of AI-generated content is still uncertain. National rules and courts can differ in detail.
Separately, the EU AI Act asks providers of general-purpose AI models to have a policy to comply with EU copyright law, including respecting rights holders' opt-outs from text and data mining, and to publish a summary of the content used for training (Article 53).
Can AI output infringe someone else's copyright?
Yes, it can. The European Parliament study notes that EU law has no general exception for AI outputs: if an output reproduces a substantial part of a protected work, sharing or selling it can infringe, even without any intent to copy. In the US, questions about training AI on copyrighted works are the subject of the Copyright Office's Part 3 report and of ongoing court cases. For everyday business text, the practical risk is lower than for images, music, code or long passages that closely follow a known source, but it is not zero.
What vendor terms usually say
Read the terms of the plan you actually use, as consumer and business terms often differ. Many business terms say the customer owns the output, or the vendor assigns whatever rights it has in it. Some vendors offer to defend business customers against third-party copyright claims about outputs, but these promises come with conditions, for example using paid business services, keeping the vendor's safety filters switched on, and not deliberately trying to produce infringing material. They often exclude outputs you have modified or combined with other content.
While you are reading, check two more points: whether the vendor may use your inputs and outputs to train its models (some business terms say it may not), and who is responsible if an output causes a problem. Ask your legal team to note the answers for each tool your people use, so everyone works from the same rules.
This is general information, not legal advice.
What to do next
- Check your AI vendor's business terms for output ownership, use of your data, and any copyright indemnity and its conditions, with your legal team.
- Where ownership matters (brand assets, publications, software), have people make and record real creative choices and edits rather than publishing raw output.
- Before publishing images, code or long text produced with AI, check it is not a close copy of an existing work, and keep a note of how it was made.
Goes deeper in our free AI Field Guide: 9. Checking AI output before you use it, 10. Data safety: what never to paste, and other rules. Ask for these guides.
Talk it through with us
Your people learn to use AI as a tool they direct and check, with tools and data rules agreed with your legal and IT teams, so the work they publish is theirs to stand behind.
Book my AI Value BriefingA conversation, not a pitch. We will not ask for confidential data. How we handle your details.How this page was made: researched and drafted with AI assistance from the sources below, each opened and checked on the date shown, and reviewed against our editorial rules.
Related questions
- U.S. Copyright Office, Copyright and Artificial Intelligence (report and guidance index) (checked 2026-10-27)
- U.S. Copyright Office, Copyright and Artificial Intelligence, Part 2: Copyrightability (January 2025) (checked 2026-10-27)
- U.S. Copyright Office, Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence (March 2023) (checked 2026-10-27)
- European Parliament, Generative AI and Copyright: Training, Creation, Regulation (study, July 2025) (checked 2026-10-27)
- European Commission, AI Act Service Desk: Article 53, Obligations for providers of general-purpose AI models (checked 2026-10-27)
- Anthropic, Commercial Terms of Service (example of vendor terms on outputs and IP indemnity) (checked 2026-10-27)
- Microsoft, Copilot Copyright Commitment (example of a vendor indemnity and its conditions, September 2023) (checked 2026-10-27)