Lawyers hit with fines after AI flubs fill their filings (nypost.com)

A wave of court sanctions is following a string of legal filings riddled with AI-generated fabrications — fake cases, bogus citations and invented authority — as attorneys increasingly rely on generative tools like ChatGPT, Microsoft Copilot and Word plug-ins (e.g., Ghostwriter Legal). Judges across multiple jurisdictions have fined lawyers for submitting hallucinated material (notable penalties include $1,000, $5,000 and an $85,000 sanction), referred some to grievance committees, and publicly rebuked claims that malware, client help, or mere unfamiliarity with AI explain the errors. Courts have called out repeated offenders and warned that excuses such as “it’s tedious to toggle programs” or “I didn’t know AI makes things up” are unacceptable. The significance for AI/ML and legal communities is twofold: it spotlights a real-world failure mode of large language models — confident but false outputs — and it underlines the necessity of human verification, clear tool provenance, and domain-specific guardrails when integrating generative AI into high-stakes workflows. Technical implications include the risk of hallucinated citations unless models are grounded to verified databases, the perils of superficial plugin integrations that obscure provenance, and escalating professional-liability and ethical consequences. Judges’ blunt admonitions — calling use of generative research “playing with fire” — signal likely tougher oversight, stricter malpractice standards, and demand for auditable, citation-accurate legal AI systems.