Background

In Advance Local Media LLC v. Cohere Inc. (2025), a group of major media publishers, including The Toronto Star and The Atlantic (the “Publishers”) accused Canadian AI firm Cohere of several acts of wrongdoing, including trademark infringement. The court in the Southern District of New York ruled in November 2025 on Cohere’s motion to dismiss the case.

The crux of the Publisher’s trademark claim was that Cohere’s AI platform allegedly hallucinated fake news articles and attributed those fake articles to registered trademarks of the Publishers. The Publishers argued this would lead users to believe that the hallucinated articles were written by, associated with, or approved by the Publishers.

Cohere argued:

1. the conduct was not commercial usage,
2. consumer confusion was unlikely, and
3. the conduct fell within nominative fair use.

The court found:

1. the Publisher’s adequately alleged there was commercial usage,
2. the Publishers adequately alleged there was a likelihood of confusion, and
3. the nominative fair use doctrine does not allow a defendant to falsely attribute its own goods to the plaintiff.

Potential Impacts on Trademarks

So far, the impact of AI on copyright law has caught the bulk of media attention. There are, as of the date of writing, more than fifty cases before courts regarding various AI-copyright matters. As such, there is more guidance in the realm of copyrights. The Cohere case joins other early cases like Getty Images v Stability AI (2025) in the UK to provide some preliminary guidance on how AI impacts trademarks.

Although the facts in Cohere are narrow and arise within a US statutory scheme, the case advances several issues that Canadian courts may ultimately confront.

The case suggests that trademark-based claims may become a viable tool alongside other claims against AI companies, especially where an AI product misidentifies authorship or creates reputational risks.

The case also suggests that the range of potential defences available to AI companies may be narrower than initially presumed.

The case suggests that AI hallucinations have the potential to cause consumer confusion, which is a central component of trademark infringement. It may not be a viable defence for AI companies to say, ‘it was the AI, not us’ – or to blame the infringement on a hallucination.

The court also seemed unpersuaded by Cohere’s fair use defence. Fair use is a related defence in both copyright and trademark matters, but with different criteria and applicability. Fair use has been a central pillar of AI company defences in the realm of copyrights. But in Cohere, in the realm of trademarks, the court seemed less convinced that fair use was a valid defence, at least within the facts of this case.

Practical Considerations

Practically, a ruling in favour of the Publishers in Cohere could have a profound impact:

1. for AI companies: guard rails may be needed to prevent offences against trademark owners,
2. for AI users: AI products may become more censored and/or expensive, and
3. for trademark owners: AI companies could be committing offences through AI products.