AI Policy & Risk

Who Owns AI-Generated Images and Text? Copyright Basics for Canada and the US

Who owns AI-generated images and text? AI-generated content copyright in Canada and the US explained, with 2026 case updates and six steps for your business.

The short answer

In the US, purely AI-generated images and text are not protected by copyright, because the law requires a human author; the Supreme Court declined to revisit that rule in March 2026. In Canada, the government's position is that authorship must come from a human exercising skill and judgment, but no court has ruled on AI output yet. For AI-generated content copyright in both countries, you can usually own the human parts you add, such as edits, selection and arrangement, and your contracts should say who owns what.

Businesses now produce a lot of content with AI: social graphics, product descriptions, blog drafts, even logos. That raises a fair question. If a competitor copies your AI-made hero image or your product copy, can you stop them? The answer depends on how much a human contributed, and it differs slightly between Canada and the US. This is general information, not legal advice.

The US position: human authorship is required

The US Copyright Office set out its view in January 2025. According to its announcement of Part 2 of its AI report, protection is available for AI-assisted work where a human makes "creative arrangements or modifications of the output, but not the mere provision of prompts." Protection is also possible where a human-authored work is perceptible in the output, such as your own photo or drawing that the AI transformed. And the Office concluded that the case had not been made for changing the law to protect purely AI-generated outputs.

The courts agree on the basic rule. In Thaler v. Perlmutter, a computer scientist tried to register artwork that he said his AI system created on its own. The DC Circuit held that the Copyright Act requires a human author, and as Holland & Knight reported, the Supreme Court declined to hear the appeal on March 2, 2026. A refusal to hear a case isn't a ruling on its merits, but it leaves the lower court's decision in place.

What remains open is the middle ground: how much human input is enough? The Copyright Office decides that case by case.

The Canadian position: likely similar, not yet tested

No Canadian court has yet decided whether an AI can be an author under the Copyright Act. The federal government's "What we heard" report on its copyright and generative AI consultation states that, under existing Canadian case law, "authorship must be attributed to a human who exercises skill and judgment." Most participants in that consultation said only AI content with sufficient human contribution should be protected, and the government said it "will examine whether any legislative or policy intervention could be made" to support human authorship.

The test case to watch is CIPPIC v Sahni in the Federal Court. It concerns an image called SURYAST that was registered in Canada in 2021 naming both a person and an AI painting app as co-authors. According to CIPPIC's case page, the registration was issued through the Canadian Intellectual Property Office's automated online system, and CIPO "does not verify ownership, or any other particulars, provided on the application form." CIPPIC argues an AI can't be an author and that the work lacks human skill and judgment. As of October 2026, the case page says the parties are waiting for the court to set a hearing date.

The practical lesson from that case: a Canadian copyright registration certificate doesn't prove a work is protected. It just means the form was filed.

See what Logo & Brand Identity would cost for your business

Answer a few questions about your project and budget and get a personalised plan, compared with typical market prices.

Get My Free QuoteAbout 3 minutes · no obligation · fixed price confirmed in writing

AI-generated content copyright: how much human input counts

The safest way to think about AI-generated content copyright is as a spectrum of human involvement. The more creative choices a person makes and can show, the stronger the claim.

Spectrum of human involvement and likely copyright protection: prompt only is likely unprotected; selecting and arranging outputs may protect the arrangement; substantial human edits may protect the edits; human-made work with AI assistance is generally protected More human creativity, stronger copyright claim Prompt only You type, AI makes the work Likely no copyright Select and arrange You choose and compose outputs Arrangement may be protected Human edits You rewrite, redraw, modify Your changes may be protected Human-made, AI-assisted AI helps with spelling, ideas Generally protected
Figure: A simplified view of how human involvement affects copyright in AI-assisted work, based on the US Copyright Office's Part 2 report. Canadian courts have not yet ruled. (Flowmatix illustration)

Some everyday examples:

  • A social graphic generated from a one-line prompt and posted as-is: probably nobody owns the image itself, including you.
  • A blog post drafted by AI and then substantially rewritten by your marketing lead: your rewrite is likely protected, even if some AI sentences aren't.
  • A brochure laid out by a designer using some AI images among photos and original text: the overall design and your text are likely protected.
  • Your own product photo that you edit with AI tools: your photo remains yours.

What the AI vendor's terms add, and don't

Most business AI terms say you own what the tool produces. Anthropic's Commercial Terms of Service, for example, state that the customer "owns its Outputs" and that Anthropic assigns to the customer "its right, title and interest (if any)" in them. Note the "(if any)". A vendor can hand over whatever rights it has, but it can't create a copyright the law doesn't recognise.

Some vendors also offer protection if a third party claims your output infringes their work. Anthropic's commercial terms include an indemnity for paid use, with exclusions, for example where you modified the output or the claim relates to your inputs. Read the exclusions before relying on it.

Six practical steps for a business

  1. Add the human layer to anything you want to protect. Edit, rewrite, combine and compose. If it matters to your brand, don't publish raw output.
  2. Keep a simple record. Save the prompt, the raw output, and your edited version with dates. That is your evidence of human contribution if it's ever questioned.
  3. Update contracts with freelancers and agencies. Ask them to disclose AI use, assign all rights they hold, and confirm they've checked outputs don't copy third-party work.
  4. Be careful with logos. A purely AI-generated logo may not be protected by copyright. Trademark protection works on different principles, so talk to a trademark professional before you invest in a brand mark. Our brand identity service page explains how we approach original marks.
  5. Don't prompt for someone else's style or characters. Ownership is one question; infringing someone else's work is another, and it's the riskier one.
  6. Label where required. Platform and EU labelling rules for realistic AI media are separate from copyright; our guide to AI video marketing covers them, and our explainer on Content Credentials shows how to attach provenance to your files.

Key takeaways

  • US law requires a human author; prompts alone don't make you one, and the Supreme Court declined to revisit this in March 2026.
  • Canada's government says authorship requires human skill and judgment, but no court has ruled on AI output; CIPPIC v Sahni is the case to watch.
  • You can usually protect the human parts: edits, selection and arrangement.
  • Vendor terms assign you whatever rights exist, which may be none for raw output.
  • This is general information, not legal advice.

Frequently asked questions

Can I copyright an image I made with an AI image generator?

In the US, not if the image comes only from your prompt; the Copyright Office says prompts alone aren't enough. If you substantially modify it or combine it creatively with other work, those human contributions may be protected. Canada has no court ruling yet.

Who owns ChatGPT or Claude output for business use?

Under most business terms, the vendor assigns its rights in the output to you. Anthropic's commercial terms, for example, say the customer owns its outputs. But if copyright law doesn't protect a purely AI-generated work, there may be no copyright to own.

Is AI-generated content copyrighted in Canada?

Unclear for now. The federal government's consultation report says Canadian law attributes authorship to a human exercising skill and judgment, and the Federal Court is considering a challenge to an AI co-authored registration. Until it rules, assume raw AI output is weakly protected at best.

Sources

  1. Copyright Office Releases Part 2 of Artificial Intelligence Report — US Copyright Office, January 29, 2025
  2. The Final Word? Supreme Court Refuses to Hear Case on AI Authorship and Inventorship — Holland & Knight, March 3, 2026
  3. Consultation on Copyright in the Age of Generative Artificial Intelligence: What we heard report — Innovation, Science and Economic Development Canada, February 2025
  4. CIPPIC v Sahni — CIPPIC, accessed October 17, 2026
  5. Commercial Terms of Service — Anthropic, effective June 17, 2025

How Flowmatix can help

We put this into practice for businesses in Toronto and around the world through our Logo & Brand Identity, content & video production services. Want to know what it would take for your business? Get a personalised quote in about 3 minutes.

Have an idea?

Ready to see what your project costs?

Answer a few questions and get a personalised plan, a comparison with typical market prices, and a clear next step. No obligation, no sales pressure.

Contact us