Magistrate Judge Lisa J. Cisneros has set a Discovery Hearing set for June 6, 2025, 9:00 AM, by videoconference. The controversy stems from the plaintiffs’ expert Dr. Ben Y. Zhao’s participation in creating the Nightshade tool, which deploys “prompt-specific poisoning attacks that corrupt a model’s ability to respond to specific targeted prompts.” “Nightshade [is] a prompt-specific poisoning attack optimized for potency that can completely control the output of a prompt in Stable Diffusion’s newest model (SDXL) with less than 100 poisoned training samples.” Zhao is also involved in creating another disruptive tool called Glaze, which he described in his declaration: “Glaze is a system designed to protect human artists by disrupting style mimicry.”
“[W]e have designed and implemented Nightshade, a tool that turns any image into a data sample that is unsuitable for model training. More precisely, Nightshade transforms images into ‘poison’ samples, so that models training on them without consent will see their models learn unpredictable behaviors that deviate from expected norms, e.g. a prompt that asks for an image of a cow flying in space might instead get an image of a handbag floating in space.” [More about these adversarial tools to poison AI data here.]
Given Dr. Zhao’s own personal adverse interest to the AI companies–the two tools are trying to defeat, if not sabotage, their models—the defendant companies seek to bar him from accessing highly confidential information related to their businesses.
Judge Cisneros appears to be contemplating whether Dr. Zhao should be disqualified and has asked for briefing on whether an alternative plaintiffs’ expert exists:
The Court has received the parties’ 300 letter brief regarding Dr. Zhao and sets a remote hearing on June 6, 2025, at 9:00am to address this dispute. The hearing will be held via Judge Cisneros’s public Zoom webinar. Plaintiffs contend that Dr. Zhao is uniquely qualified to serve as an expert and they would be severely prejudiced by his disqualification. ECF No. 300 at 4. Defendants contend that “Plaintiffs have already offered” an alternative AI expert “to whom Defendants did not object to disclosing the very same Highly Confidential information” and that Plaintiffs’ claims of prejudice are thus exaggerated. Id. at 6.
Plaintiffs shall file, by June 2, 2025, supplemental briefing not to exceed three pages responding to Defendants’ contention that Plaintiffs have already “identified” a “non-conflicted alternative” to Dr. Zhao and thus Plaintiffs would not be severely prejudiced by Dr. Zhao’s disqualification. Such supplemental briefing shall address whether Dr. Zhao possesses unique knowledge within the field of AI image generation that make him “better suited than any other expert[.]” GPNE Corp. v. Apple Inc., 12-cv-02885, 2014 WL 1027948, at *1 (N.D. Cal. Mar. 13, 2014) (requiring the proponent of an expert to make a “particularized showing that the expert had unique qualifications that other experts could not provide”). Defendants may file a response, not to exceed three pages, by June 4, 2025. (This is a text-only entry generated by the court.
On his webpage, Dr. Ben Zhao said: ” Since 2022, I work primarily on adversarial machine learning and tools to mitigate harms of generative AI models against human creatives in different industries.”
Here are videos of interviews with Dr. Ben Zhao:
Related Stories