The list of cases that target the developers/operators of generative artificial intelligence (“AI”) platforms, most of which tend to focus heavily on copyright infringement and Digital Millennium Copyright Act (“DMCA”) claims, keeps growing. Among a number of newly filed cases is one that was filed against Microsoft and OpenAI in the U.S. District Court for the Southern District of New York on April 30. The “publisher plaintiffs,” which include the companies behind the New York Daily News, the Chicago Tribune, the Orland Sentinel, and the Denver Post, among others, set out an array of copyright-centric causes of action, such as direct, vicarious, and contributory copyright infringement.
From a copyright perspective, this case – and others like it – centers on two key issues: One at the input/training stage and one that has to do with AI-generated output. Primarily, the publisher plaintiffs take issue with OpenAI and Microsoft’s allegedly unauthorized use of the newspapers’ content to train the large language models that power their generative AI offerings, including ChatGPT and Copilot. In terms of output, the plaintiffs claim that OpenAI and Microsoft have run afoul of copyright law by providing users of their generative AI tools with “content that is identical to, or a slightly masked version of, the newspapers’ content.”
Primarily, “This lawsuit is about how Microsoft and OpenAI are not entitled to use copyrighted newspaper content to build their new trillion-dollar enterprises, without paying for that content,” the plaintiffs assert. At the same time, the case is about how ChatGPT and Copilot “display copies or derivatives of the publisher [plaintiffs’] works” – including “nearly the entire text of” the plaintiffs’ original articles – in response to user prompts. And in at least some cases, the plaintiffs claim that ChatGPT and Copilot provide such “verbatim” responses as swiftly as a day after the original articles are published.
> In terms of the DMCA issues, this case and other similar ones accuse the defendants of removing copyright management information – such as “a copyright notice, authors’ names, publisher’s name, title and other identifying information, terms and conditions of use, and identifying numbers or symbols referring to the copyright-management information” – when building the training datasets containing copies of the publishers’ works.
Putting copyright and DMCA claims aside, the news publishers’ case against OpenAI and Microsoft is an interesting one, in part, because of its inclusion of trademarks claims. This is noteworthy, as trademark causes of action have been far less common (than copyright ones) in generative AI cases. In the case at hand, the plaintiff publishers wage a federal trademark dilution claim (15 U.S.C. § 1125(c)), as well as one under New York state law (N.Y. Gen. Bus. Law § 360-l).
The crux of these claims is that in addition to “copying, reproducing, and paraphrasing the publishers’ content without consent or compensation,” the defendants – by way of their generative AI models – are “also causing the publishers commercial and competitive injury by misattributing content to the [them] that the [they] did not, in fact, publish” and using their branding on the erroneous GPT-generated outputs. For example, the plaintiffs claim that “in response to a [user] prompt requesting an informative essay about Denver newspapers’ reporting that smoking cures asthma,” one of the defendants’ GPT models “completely fabricated that the Denver Post published research and medical observations that smoking can be a cure for asthma.”
As an “actual and proximate result of the unauthorized use of [their] trademarks” in connection with such inaccurate reporting (or “hallucinations” as they are referred to in AI speak), the publisher plaintiffs maintain that they “have suffered and continue to suffer harm by, among other harms, harming their reputation for accuracy, originality, and quality, which has and will continue to cause them economic loss.”
In short: The plaintiffs claim that OpenAI and Microsoft’s “unauthorized use of [their] trademarks” to identify “lower-quality and inaccurate writing dilutes the quality of the [marks] by tarnishment, in violation of 15 U.S.C § 1125(c),” and “dilute[s] the distinctiveness of the trademarks, as well as injures [their] business reputation” in violation of New York General Business Law § 360-l.
Not the very first case to bring trademark claims into the fold: In February 2023, Getty Images filed suit against Stability AI, the company behind AI-powered image generator Stable Diffusion, setting out copyright claims, as well as trademark infringement and dilution ones. On the latter front, Getty has argued that Stable Diffusion’s frequent generation of output bearing a modified version of the Getty Images watermark creates “confusion as to the source of the images and falsely impl[ies] an association with Getty Images” – hence, its infringement cause of action.
In addition to causing confusion, Getty also asserts that “while some of the output generated through the use of Stable Diffusion is aesthetically pleasing, other output is of much lower quality and at times ranges from the bizarre to the grotesque.” As such, Stability AI’s “incorporation of Getty Images’ marks into low quality, unappealing, or offensive images dilutes those marks in further violation of federal and state trademark laws.”
While Stability AI has sought to escape Getty’s claims, the case is still underway before the U.S. Districr Court for the District of Delaware.