California is on the verge of becoming one of the first states to directly regulate how generative AI may be used by attorneys and arbitrators in legal practice. A bill that cleared the state legislature now awaits action by Gov. Gavin Newsom, putting the legal industry on notice that AI governance is moving beyond ethics opinions and court standing orders into statute.
While courts and bar regulators around the country have already warned lawyers about AI-related risks — including hallucinated citations, confidentiality concerns, and inadequate supervision of automated work product — this measure is significant because it would create a state-level framework specifically aimed at participants inside the legal system. That makes it more than a general AI policy debate. It is a legal operations issue for anyone practicing in California.
For litigators, the practical implications are immediate. Generative AI is now routinely used for research summaries, drafting assistance, document review, and internal case analysis. A California law focused on attorney use of AI could affect disclosure obligations, diligence standards, and the extent to which lawyers must review or verify machine-generated output before relying on it in pleadings, briefs, discovery responses, or settlement discussions. Even if the bill does not prohibit AI use, it signals that unsupervised or opaque use will face closer scrutiny.
The inclusion of arbitrators is especially notable. Arbitration has often moved faster than courts in adopting technology, but with less public visibility. Regulating AI use by neutrals raises questions about fairness, transparency, and whether parties should know when AI tools influence procedural decisions, legal analysis, or award drafting. Companies that rely heavily on arbitration clauses may want to watch this closely, particularly if they manage large volumes of employment, consumer, or commercial disputes in California.
For in-house counsel and compliance teams, the bill is another reminder that enterprise AI policies must account for legal-specific workflows. It may no longer be enough to have a broad corporate AI policy. Legal departments may need separate rules governing approved tools, human review requirements, confidentiality protections, vendor diligence, and documentation of how AI-assisted work is created and checked.
If signed, the legislation would likely influence policy discussions well beyond California. Other states, bar regulators, and courts are watching the same issues, and California often serves as an early model for technology regulation. Legal professionals should treat this development as a sign that AI governance in law is entering a more formal phase — one where internal guardrails, training, and defensible review practices may become just as important as the tools themselves.