Think Before You Prompt: Copyright Risks in the Age of Generative AI

May 13, 2025

As generative AI tools like ChatGPT, Perplexity, Claude, Gemini, and others become embedded in everyday business use, many companies see new opportunities for productivity, creativity, and speed. But the U.S. Copyright Office’s recent report [1] offers a critical reminder: using AI may come with copyright risk—some of it yours.

The big question for companies isn’t just whether AI companies like OpenAI or Stability AI could be liable for using copyrighted material to train their models. It’s whether you, the user, might face risks too.

First considering the infringement by the AI Tools themselves, the Copyright Office has foreshadowed its analysis, which appears to create significant risk to AI tools in training models on copyrighted material, although actual infringement (or Fair Use) will depend on the specific facts and circumstances.

Is use of Copyright Works by AI Tools Considered Fair Use?

1. The determination of whether the use of copyrighted works in AI training constitutes fair use is a nuanced, fact-specific inquiry guided by the four statutory factors codified in Section 107 of the Copyright Act. In its recent report, the Copyright Office concludes that several stages in the development of generative AI involve using copyrighted works in ways that implicate the owners' exclusive rights. The key question then is whether these acts of prima facie infringement can be excused as “fair use” under the Copyright laws.[2] Various uses of copyrighted works in AI training are likely to be transformative, which them would qualify them as “fair use”, but the extent to which they are fair depends on multiple factors such as the purpose, source, and controls on outputs.

With that said, the Copyright Office has foreshadowed the risks to AI companies training their tools on copyrighted material that is used for commercial purpose:

“[t]he copying of expressive works from pirate sources in order to generate unrestricted content that competes in the marketplace, when licensing is reasonably available, is unlikely to qualify as fair use.”[3]

While there are some uses--such as noncommercial research or analysis that do not enable portions of the works to be reproduced in outputs—are likely to be fair, the Copyright Office has made clear that commercial uses that could impact competition in the marketplace, likely are not.

So, how does all of this effect you, as the user of generative AI tools? Good question.

Key Copyright Risk Areas for AI Users

1. Direct or Contributory Infringement by Users
If a user prompts an AI tool in a way that results in the generation of content that substantially copies protected works—think a nearly identical version of a famous artist’s song or a popular author’s storyline—the user could be exposed to claims of direct or contributory copyright infringement. This is especially risky if the content is then shared publicly or commercialized.
2. No Clear Safe Harbor
Contrary to what some might assume, end users aren’t automatically shielded by the AI company's actions. The Copyright Office makes clear that users could still be held responsible if they knowingly use or distribute infringing output from these tools—even if they didn’t train the AI or write its code.
3. "Substantial Similarity" Still Applies
Legal standards like “substantial similarity” and “access to the original work” still govern infringement. If an AI-generated image, song, or piece of text is too close to an existing copyrighted work, it may not matter that the AI did the writing—you used it and published it.

"Whether a model's weights implicate the reproduction or derivative work rights turns on whether the model has retained or memorized substantial protectable expression from the work(s) at issue ... Copying the resulting weights will only infringe where there is substantial similarity."[4]

What Should Executives and Teams Do?

Don’t assume outputs are “original.” AI doesn’t “create” in the human sense—it blends, predicts, and remixes. That beautiful image or slick marketing copy could echo or incorporate a protected work.
Treat generative content like third-party content. Use the same diligence you’d apply if outsourcing to a freelancer: check licenses, vet for originality, and avoid content you can’t trace.
Create policies for AI use. Make sure employees understand the risks and know when to escalate content for legal review before it’s published.
Negotiate Copyright Warranties and Indemnities. Where you can, for example for larger enterprise engagement with AI companies, try to include in the contract explicit warranties of non-infringement, lawful acquisition of data for LLM training purposes, and seek full indemnification for all claims related to your use of output and based on the AI tool/LLM training on third-party content.
Require Training Data Disclosure / Transparency, Output Filtering, Remediation Procedures, and Audit Rights. Again, where able, incorporate provisions in the contract to include (i) an affirmative requirement to disclose on what data the LLM trains, including whether copyrighted works were used, (ii) technical filtering to identify potentially infringing output and prevent its publication by the AI Tool, (iii) procedures for notifying you if any infringing content is identified, and include prompt take down obligations (including retraining upon model updates), and (iv) the ability to audit the AI provider’s compliance with these obligations.
Stay informed. Courts haven’t resolved all questions yet, but early cases (like those against AI developers by artists and authors) suggest that downstream users could be dragged into litigation.

Generative AI can supercharge your team—but it also comes with risks. If you use these tools without oversight, your company could inherit not just creativity, but copyright claims. However, with any new tech, the key is not to avoid adoption out of fear, but evaluate the benefits and the risks. Foresight and preparation are critical to responsibly incorporating such tech into your business.

Breaking News . . .

Now, with all of that said, after release of the report, the head of the U.S. Copyright Office, Register of Copyrights, Shira Perlmutter, was terminated . . . so, all of this could change!

Stay tuned!

Think Before You Prompt: Copyright Risks in the Age of Generative AI

Discussion about this post