Introduction
The legal proceedings in the Andersen v Stability AI case highlight the complexities of intellectual property rights in the realm of artificial intelligence. This case [3], involving visual artists and AI companies, centers on allegations of copyright infringement and the contentious role of expert witness Dr Ben Y Zhao.
Description
Magistrate Judge Lisa J Cisneros held a Zoom hearing on June 6, 2025 [3], in the Andersen v Stability AI case [2], which involves a class action initiated by visual artists against Stability AI, Midjourney [1], and DeviantArt for copyright infringement [1], DMCA violations [1], and trade dress claims [1]. While the court has dismissed several claims [1], it has retained the direct copyright infringement and trademark claims [1], making the involvement of expert Dr Ben Y Zhao particularly significant.
Dr Zhao [1] [2] [3] [4], a computer science professor at the University of Chicago [2], is known for co-creating the Nightshade tool [3], which executes prompt-specific poisoning attacks that can significantly impair the performance of AI models [3], particularly targeting Stable Diffusion’s latest model (SDXL) with minimal poisoned training samples [3]. He is also associated with another tool [3], Glaze [3] [4], designed to protect human artists from style mimicry [3]. The plaintiffs argue that Dr Zhao’s unique qualifications are essential for supporting their claims, while the defendant AI companies contend that his involvement poses a competitive risk and creates a conflict of interest due to his development of public-facing adversarial software.
The defendants filed a motion to prevent Dr Zhao from accessing their highly confidential datasets [2] [3], asserting that this access could harm their business interests. Defense attorney Joe Gratz cited case law supporting the exclusion of experts from confidential information if there is a potential for misuse [2]. While he expressed willingness for Dr Zhao to testify [2], he opposed granting him access to sensitive datasets that could enhance the effectiveness of his tools [2].
In contrast [2] [4], Joshua Stein [2], representing the plaintiffs [2], argued that the cited cases did not pertain to academics who are not direct competitors [2]. He defended Nightshade as a tool that prevents the misappropriation of data rather than one that undermines the defendants’ interests. The plaintiffs contended that denying Dr Zhao access to the materials necessary for analysis would cause them significant prejudice. They also proposed an alternative AI expert [3], which the defendants claimed undermined the plaintiffs’ assertions of prejudice [3]. The discussion included the availability of alternative experts [2], with Judge Cisneros questioning the plaintiffs’ claims of being unable to find another expert not associated with such tools [2]. Gratz maintained that no conditions could sufficiently mitigate the risks posed by Dr Zhao’s access to the datasets [2].
By the conclusion of the hearing [2], Judge Cisneros indicated the need for further deliberation and may require additional briefing [2], postponing the decision on Dr Zhao’s disqualification as the plaintiffs’ expert until June 10, 2025 [2]. The plaintiffs are required to submit supplemental briefing addressing the defendants’ argument regarding the alternative expert by June 2, 2025 [3], with the defendants having the opportunity to respond by June 4, 2025 [3]. The case centers on the contentious issue of whether Dr Zhao should be granted access to the confidential source code of the AI companies involved [3].
Conclusion
The outcome of this case could have significant implications for the intersection of AI technology and intellectual property law. The decision on whether Dr Zhao can access confidential datasets will not only impact the current litigation but may also set a precedent for how expert witnesses are utilized in similar cases involving proprietary AI technologies.
References
[1] https://www.mckoolsmith.com/newsroom-ailitigation-26
[2] https://chatgptiseatingtheworld.com/2025/06/10/magistrate-judge-cisneros-finds-difficult-question-whether-dr-ben-zhao-should-be-allowed-to-see-confidential-business-information-of-ai-image-generators-despite-developing-tools-to-poison-them/
[3] https://chatgptiseatingtheworld.com/2025/06/04/hearing-set-to-determine-whether-plaintiffs-expert-dr-ben-y-zhao-who-created-tools-to-poison-ai-models-data-should-be-allowed-to-see-highly-confidential-material-of-stability-ai-or-should-zhao/
[4] https://www.mckoolsmith.com/newsroom-ailitigation-25
