Courts Grapple with the Challenges of AI-Generated Cases in Recent Litigation

Denver, Colorado – The use of artificial intelligence (AI) in legal research has raised concerns about the accuracy and trustworthiness of AI-generated information in court filings. Last year, we highlighted the risks associated with generative AI tools, including the unintentional citation of non-existent cases by lawyers and the subsequent imposition of limits on AI use by courts. These trends have continued, with recent cases involving lawyers who relied on AI tools to draft motions and submitted AI-generated legal cases without verifying their accuracy.

In one instance, a Colorado practitioner drafted a motion using the assistance of ChatGPT, an AI tool, and included citations to cases that were later found to be fictitious. The lawyer initially blamed a legal intern for the error, but later admitted to the unverified use of ChatGPT. As a result, the lawyer was suspended from practice for a year. Similarly, former Trump counsel Michael Cohen submitted AI-generated legal cases from Google Bard to his lawyer, who included them in a federal court filing. Cohen claimed that he mistakenly believed Google Bard to be a reliable tool for legal research. The court has yet to decide on potential sanctions in this case.

To address these issues, a growing number of judges have issued standing orders requiring litigants to disclose if their documents were AI-generated. The Fifth Circuit is considering implementing a court-imposed AI accuracy check. Failure to disclose AI use or misuse may result in sanctions. A recent case in the Second Circuit involved a lawyer who cited a phantom case generated by ChatGPT in a brief. The court emphasized that existing rules already require claims and legal contentions to be properly supported and legally tenable, and referred the lawyer to the Court’s Grievance Panel.

To avoid these pitfalls, lawyers are encouraged to verify the information provided by generative AI tools, promptly notify the court and relevant parties if a phantom case is cited, and comply with applicable policies on AI use in legal research.

Moving on, a Fourth Circuit appeal in the case of Peninsula Pathology Associates v. American International Industries raises questions about whether courts should protect made-for-litigation studies. The appeal asks whether a party can seek discovery into a study suspected of being a ploy to present other plaintiffs’ allegations before a jury. The case involves a plaintiff alleging that exposure to asbestos in cosmetic talc products caused them to develop mesothelioma. The plaintiff’s experts relied on a study based on talc-asbestos lawsuits which reported cases of mesothelioma allegedly caused by cosmetic talc. The defendant subpoenaed the experts’ employer for the names of the individuals in the study, but the court quashed the subpoena citing minimal relevance and privacy concerns.

Another significant development comes from the Pennsylvania Supreme Court, which held that evidence of a manufacturer’s compliance with industry or government standards is inadmissible as a defense in design defect cases under the risk-utility test in strict liability. In Sullivan v. Werner Co., the court concluded that such evidence diverts attention from the product’s attributes and does not prove any characteristic of the product. Although compliance evidence is not relevant to design defect claims, noncompliance evidence may still be admissible.

In environmental news, a draft Cancer Incidence Study (CIS) by the Agency for Toxic Substances and Disease Registry (ATSDR) will not be produced in the Camp Lejeune water litigation. The plaintiffs requested the production of the draft CIS, which was still undergoing peer review, but the court denied the motion, citing the deliberative process privilege. The court recognized the public interest in maintaining the integrity of ATSDR’s peer review procedure and avoiding premature release of incomplete scientific products.

The Environmental Protection Agency (EPA) released its Second Annual Progress Report on per- and polyfluoroalkyl substances (PFAS), outlining milestones achieved and forthcoming regulatory activities. EPA expects to finalize enforceable national drinking water standards for six PFAS in early 2024 and to list PFOA and PFOS as hazardous substances under CERCLA. The report also mentions other initiatives, such as including PFAS as a pollutant under its Air Emissions Reporting Requirements and inviting public comments on risk assessments for PFOA and PFOS in biosolids.

Finally, a review of the Judicial Panel on Multidistrict Litigation’s (JPML) activities in product liability litigation reveals that product liability cases continue to comprise a significant portion of active MDLs. While the overall number of MDLs has declined, the JPML continues to grant petitions for centralization, citing convenience for parties and witnesses, avoidance of duplicative discovery, prevention of inconsistent pretrial rulings, and streamlining of pretrial proceedings. The JPML is expected to consider more MDL petitions in the coming year.

In conclusion, the use of AI in legal research, challenges to the protection of made-for-litigation studies, the inadmissibility of compliance evidence in design defect cases, developments in PFAS regulation, and the trends in multidistrict product liability litigation are all significant and evolving issues in the legal landscape. Lawyers and courts must carefully navigate these challenges and adapt to a rapidly changing legal environment.