State-bar AI ethics opinions: a 2026 explainer

The ABA gave us the national framework in Formal Opinion 512. But you're admitted in a state, and several state bars have now issued their own AI guidance. The themes rhyme — competence, confidentiality, candor, supervision, honest billing — but the specifics are worth knowing. Here's the lay of the land in five key jurisdictions, in plain English. Always read the source guidance for your own jurisdiction; this is an explainer, not advice.

California

The State Bar of California issued Practical Guidance for the Use of Generative AI in the Practice of Law (2023) — one of the earliest. It stresses confidentiality (don't input client information into a tool that lacks adequate confidentiality protections), competence, and not charging for time the AI saved. See our California page.

New York

The New York State Bar Association's Task Force on Artificial Intelligence published a substantial report and guidelines (2024) covering risks, benefits, and recommendations for responsible adoption. The throughline is the same duties — competence, confidentiality, supervision — applied to AI. See our New York page.

Florida

The Florida Bar issued Ethics Opinion 24-1 (2024) on lawyers' use of generative AI, addressing confidentiality, oversight of the work product, billing, and advertising. It's one of the more concrete state opinions. See our Florida page.

Texas

The State Bar of Texas formed a Taskforce for Responsible AI in the Law (TRAIL) to study AI's impact on the profession, and in February 2025 its Professional Ethics Committee issued Opinion 705: lawyers must understand AI tools, protect confidential information, verify output, and charge reasonable fees — not billing hourly clients for time AI saves. See our Texas page.

Illinois

The Illinois Supreme Court adopted a Policy on Artificial Intelligence (effective 2025) addressing AI use in the courts and by practitioners. See our Illinois page.

The common thread — and what to do Monday

1. Protect confidences. Use a tool with a real no-training posture; don't paste client facts into a free consumer tier.
2. Verify output. Validate every citation against a real database before filing.
3. Bill honestly. Charge time spent, not time saved.
4. Disclose when reasonably necessary. An engagement-letter AI clause is a common starting point, though boilerplate alone may not be enough where informed consent is required.

How Draftiro maps to the guidance

Draftiro supports these duties with detected supported-case citation checks when CourtListener is available, statutes flagged "unchecked," an audit-logged conflict check, and draft AI-assisted time entries for case-scoped work that attorneys can review, edit, or delete before billing. It does not determine compliance. We track these opinions on our AI ethics page.