Holding / Framework
Court found the false citations 'embarrassing and certainly negligent' but declined to impose sanctions, finding neither Cohen nor Schwartz acted in 'bad faith.' Holding establishes the GOOD-FAITH FRAMEWORK for AI-hallucination cases: where the filer demonstrates a genuine misunderstanding of the AI tool's capabilities (rather than reckless disregard), Rule 11 sanctions may not lie. This is a doctrinal counterpoint to Mata v. Avianca's subjective-bad-faith finding on similar facts; together they bracket the scienter spectrum for AI-citation misconduct.
Triggering Conduct
Cohen included three fabricated case citations in a brief seeking early termination of supervised release. Citations were generated by Google Bard, which Cohen reportedly believed was a 'super-charged search engine' rather than a generative-text service. Cohen forwarded the citations to Schwartz, who incorporated them into the filing without independent verification.
Sanctions / Disposition
NONE imposed. Court declined to sanction either Michael Cohen (former Trump attorney; pro se litigant in his post-prison supervision matter) or his attorney David M. Schwartz.
Primary Source
https://www.courthousenews.com/judge-wont-sanction-michael-cohen-over-ai-generated-fake-legal-cases/ · secondary_aggregator (Courthouse News / ABA Journal); primary docket pull pending
Tags
2d_circuit, hallucinated_citations, no_sanctions, good_faith_framework, scienter, framework_setter, criminal_supervised_release