Guardrails to Prevent Bias and Breaches
The committee flagged several dangers of AI use in legal contexts, including the risk of fabricated information, bias, and exposure of confidential materials. Judges are strictly barred from using AI to render decisions, and nonjudicial employees may not employ the technology in ways that violate ethical obligations.
All confidentiality and data security rules that govern court records now explicitly extend to AI systems. The policy warns that information entered into public AI models should be treated as if it will become instantly public.
Approved Tools and Banned Practices
Among the greenlit tools are Microsoft Azure AI Services, Microsoft 365 Copilot, GitHub Copilot for Business or Enterprise, and Trados Studio. The free version of OpenAI’s ChatGPT is also permitted—though paid subscriptions remain off-limits.
Any judicial officer facing ethical uncertainties related to AI use is encouraged to consult the Advisory Committee on Judicial Ethics, according to Friday’s press release.