Why my Midjourney AI case is one to watch’: Cleary partner on ‘art style’ IP

Fresh from her courtroom win for Meta, Angela Dunning is primed to offer unique insights to delegates at the upcoming WIPR Trademark and Brand Protection Summit.

Cleary Gottlieb partner Angela Dunning knows a thing or two about the rapidly-evolving intersection between trademark law and artificial intelligence (AI).

Dunning is currently defending AI developer Midjourney against putative copyright class actions challenging the development, training, and output of those models.

It is one of several reasons she is ideally placed to speak on the panel, The growing AI threat in trademark infringement at WIPR’s Trademark and Brand Protection Summit in late October.

Last June, she secured a notable summary judgment win for Meta in Kadrey v Meta, one of the first copyright cases involving the use of literary works to train generative AI.

In that case, US District Judge Vince Chhabria held that Meta’s use of works by 13 prominent authors to train its Llama models constituted ‘fair use’ and that the plaintiffs had failed to show market harm.

With almost 25 years of experience litigating copyright, trademark, trade secrets, right of publicity, and unfair competition/false advertising cases, Dunning also oversees complex commercial disputes for major global companies.

In addition to her win for Meta, she has substantial experience in the federal appellate courts, including successful arguments before the US Court of Appeals for the Ninth Circuit in precedent-setting IP cases such as Elliott v Google, Naruto v Slater, and Close v Sotheby’s.

Her upcoming panel at the San Francisco event is set to explore the pivotal issues affecting brand owners, including how to create new policies and processes to take down counterfeiters who use AI to infringe at scale, as well as understanding and acting on the increased risk of fraud and damage to brands from sophisticated AI infringement.

It will also look to the future, examining the likely next steps for brands, platforms, and infringers when grappling with AI.

Dunning sat down with WIPR to explore why, when it comes to infringement, AI can often be more of a help than a hindrance to brand owners.

What has been the most surprising or challenging case you’ve encountered related to AI and trademarks?

My firm and I are currently defending a putative class action, Andersen v Stability AI, on behalf of Midjourney, a leading AI image generator.

One of the claims against Midjourney is for vicarious trade dress infringement, premised on allegations that its model can be used to generate images in particular artists’ styles, which allegedly constitute their trade dress.

Art styles are not protected by copyright, and there are significant hurdles to establishing protectability as trade dress as well.

However, the ability of an artist to lock up exclusive rights to an art style as trade dress raises significant implications for the future development and evolution of art and art styles, making this claim one to watch as the case proceeds.

At present, are bad actors using AI to deliberately replicate or spoof trademarks at scale? And if so, to what extent?

The ability to generate infringing articles using protected trademarks existed long before AI and will persist irrespective of the introduction of AI or other tools for generating new content.

While AI may enable infringers to more easily ideate and create modified versions of marks, the methods of detection and enforcement should remain the same.

How effective are current detection systems in identifying AI-generated counterfeit branding or logos?

Whether a mark is counterfeit or infringing turns on a variety of factors, including the similarity of the mark in sight, sound, and meaning to the protected trademark. These factors are unaffected by the manner of creation of the alleged counterfeit.

Do current trademark laws adequately cover AI-generated infringement?

In my view, yes.  In all cases, the infringement inquiry will turn on how close the counterfeit is to the original trademark, the relatedness of the parties’ goods and services, the strength of the parties’ respective marks, any evidence of actual confusion, and a list of other factors that allow courts to assess the likelihood of consumer confusion.

Should the liability lie with the AI developer, user, or platform when infringement occurs?

As always, anyone launching a product or service under a brand or mark—irrespective of how that mark was conceived or created—should take care to ensure that the mark they have selected does not infringe the rights of another under a registered trademark or the common law.  Appropriate clearance work by qualified counsel should minimise this risk.

What role can AI itself play in detecting trademark misuse or infringement?

AI tools can be great at identifying potential infringing uses on the internet: ideating on new, non-infringing logos; comparing a proposed mark to those registered with the US Patent and Trademark Office; and even producing an assessment of the factors affecting the likelihood of confusion.

Deployed thoughtfully, and so long as all outputs are reviewed for accuracy by human lawyers, AI can provide a great starting point for detecting and policing, or avoiding, infringement.

How do you foresee the landscape of trademark enforcement evolving over the next 5–10 years with the rise of AI?

AI tools, properly deployed, should make it easier for brand owners to discover and police infringement of their marks.

WIPR Trademark and Brand Protection Summit takes place on 28-29 October in San Francisco.

Did you enjoy reading this story?  Sign up to our free daily newsletters and get stories sent like this straight to your inbox