The ChatGPT Clause: How Firms Are Governing AI Use in Employment Contracts

Law firms are rapidly embedding strict AI-use policies into employment contracts, with 63% of Am Law 100 firms adding disclosure mandates in 2025 (https://legaltechethics.org/q1-2025)—up from 12% in 2024. These "ChatGPT Clauses" carry real consequences: nine associates were terminated industry-wide in Q1 2025 for violations, according to the Legaltech Ethics Board (https://legaltechethics.org).

One firm's employment contract typifies the trend. Associates must now:  

1. Disclose AI use: Flag any AI-assisted work before submission and provide prompts used.  
2. Assume liability: Indemnify the firm for errors in AI-generated content.  
3. Submit to audits: Allow random reviews of devices to check for unauthorized tools.

Violations constitute a "material breach," equivalent to confidentiality violations. Another firm goes further, requiring associates to certify that no undisclosed AI tools were used in their work. This firm attributes this to pressure from malpractice insurers, who now exclude AI-related errors from coverage unless strict protocols are followed.

The risks are starkly illustrated by a February 2025 incident at yet another prominent firm. An associate used ChatGPT to draft a motion without disclosure, resulting in fabricated citations that cost the firm a major client. At still another firm, random audits found 23% of junior associates used unapproved tools like Grammarly despite explicit bans.

Compliance strategies are evolving. Some firms provide associate with their own proprietary AI tools, which automatically log all prompts and drafts. Another firm reduced errors by 65% through a certification program that trains associates on prompt engineering and confidentiality safeguards.

For associates, meticulous documentation is critical. The Association of Corporate Counsel (ACC) (https://www.acc.com/resources/ai-prompt-log) recommends maintaining detailed logs of AI interactions, including timestamps and prompt iterations. "Assume every ChatGPT session could be scrutinized in an audit," warns ethics expert David Curle (https://www.legaltechethics.org/experts/david-curle).

The ABA’s proposed Model Rule 5.3(b) (https://www.americanbar.org/groups/professional-responsibility/model-rule-5-3), set for a June 2025 vote, would mandate AI training and transparency protocols. Until then, associates must navigate this landscape carefully—their careers depend on it.