Responsible AI — Intellectual Property & Copyright Risk

An Employee Uses an AI Tool to Write a Marketing Campaign. The Output Is Polished and On Brand — but a Colleague Notices It Closely Mirrors a Competitor’s Published Copy. The Deadline Is Tomorrow. Can They Use It?

A real generative AI and intellectual property scenario — with three decision options and the right answer.

The Situation

A marketing manager at a software company needs to produce copy for a major product launch campaign by tomorrow morning. Under time pressure, they use an AI writing tool to generate taglines, email subject lines, and a landing page hero section. The output is sharp, polished, and well aligned with the brand voice. They share it with their team for a final review before publishing.

A colleague on the team immediately recognizes that the landing page copy — specifically a three-sentence value proposition block — is strikingly similar to a competitor’s published website copy, down to several distinctive phrases and the structural ordering of benefits. The AI has clearly been trained on web content that includes the competitor’s site. The marketing manager says, “I didn’t copy anything — the AI wrote it,” and wants to publish on schedule.

Choice APublish on schedule. The AI wrote the copy — the company didn’t plagiarize anything. The similarity is coincidental, and the competitor would have to prove intent to pursue a copyright claim.

Choice BDo not publish the flagged content. Rewrite the similar sections in the original language before the launch — or delay the launch if there isn’t time to rewrite properly. Flag the situation to Legal before publishing anything.

Choice CChange a few words in the flagged section to make it sufficiently different from the competitor’s copy, then publish on schedule.

The Right Call

Choice B — Do not publish. Rewrite the flagged content and involve Legal.

Choice A publishes with documented knowledge of a potential infringement — the colleague flagged it before publication, creating a record that the team knew and published anyway. That’s the worst legal position. Choice C — changing a few words — is unlikely to be sufficient. Copyright protection extends to the expression and structure of the original work, not just the exact words. A substantially similar structure with a few substituted synonyms may still constitute infringement. Legal review is the only way to know whether the rewrite is sufficient. The deadline pressure is real but it does not change the legal analysis.

Copyright infringement does not require intent — it requires substantial similarity.

The standard for copyright infringement is whether the infringing work is substantially similar to the protected expression in the original work — not whether the infringer intended to copy. An organization can infringe a competitor’s copyright through AI-generated content without any individual employee having copied a single sentence. The AI’s training data is the source of the similarity, but the organization that publishes the output is responsible for it.

Publishing with documented knowledge of the issue significantly worsens the legal position.

When the team publishes after a colleague flags a similarity, there is now a written record — in team messages, emails, or meeting notes — that the organization was aware of the potential infringement before publication. This matters enormously in copyright litigation. Willful infringement — infringement with knowledge — carries substantially more serious statutory damages than innocent infringement. The colleague’s flag is both an ethical obligation and a legal liability management issue.

Generative AI and copyright law are actively evolving areas, making legal review essential.

Courts are actively litigating whether AI-generated content that closely mirrors training data constitutes infringement, who holds liability — the AI developer, the user, or both — and whether AI output can be copyrighted at all. The marketing team is not qualified to assess the legal risk of publishing substantially similar AI-generated content without Legal’s input. The cost of a brief Legal review is orders of magnitude lower than the cost of defending a copyright infringement claim.

Can AI-generated content infringe copyright if the user didn’t copy anything?

Yes. Copyright infringement requires access to the protected work and substantial similarity — not intentional copying. Generative AI models are trained on large datasets that include copyrighted works. When a model produces output that is substantially similar to a specific protected expression in its training data, publishing that output may constitute infringement regardless of the user’s intent. This is an area of active litigation — several major cases are currently working through courts that will clarify the liability framework for AI-generated content.

What practical steps can marketing teams take to reduce AI copyright risk?

Best practices include: using AI as a drafting aid rather than a final output source, requiring human review of all AI-generated content before publication, running outputs through similarity-checking tools before publishing, establishing a review protocol for any content that a team member flags as potentially similar to existing published work, and involving Legal in reviewing the organization’s AI content policy. Organizations can also look for AI tools that offer IP indemnification — some enterprise AI agreements include contractual protection against copyright claims arising from tool outputs.

What is the difference between willful and innocent copyright infringement?

Under the Copyright Act, statutory damages for innocent infringement can be as low as $200 per work. Statutory damages for willful infringement — where the infringer knew or should have known the conduct constituted infringement — can reach $150,000 per work. The distinction between innocent and willful often turns on whether the organization had notice of the potential infringement before publication. A team member flagging similarity before publication is exactly the kind of notice that can convert an innocent infringement into a willful one if the organization publishes anyway.

How to Use This Scenario in Training

Recommended for marketing, communications, content, and product teams that regularly use generative AI tools. The key recognition skills are: understanding that “the AI wrote it” does not eliminate copyright liability, understanding that flagging similarity before publication is both an ethical and legal obligation, and understanding that publishing with knowledge of a potential issue converts innocent infringement into willful infringement.

This scenario captures one of the most important training principles behind the Decision Readiness Engine™: the colleague who flagged the similarity had a recognition moment. The question is whether the team paused on it — or rationalized past it because the deadline was real and the liability felt abstract. That decision moment is what this scenario is designed to train.

More Responsible AI Scenarios

Want AI Copyright Scenarios in Your Compliance Program?

Xcelus builds responsible AI training covering generative AI copyright risk, IP policy, and the review obligations that protect organizations from AI-generated infringement claims.

View the Compliance Reinforcement Kit →
Contact Xcelus