not agreeing with each other

Every state bar association is making its own call on AI ethics -- and they're not agreeing with each other. Some states demand full disclosure every time you touch a generative AI tool. Others haven't said a word. If you're practicing across jurisdictions, that patchwork isn't academic. It's a malpractice trap.

We've mapped every state's position on AI ethics for lawyers into one reference. This is the guide managing partners are bookmarking because the alternative -- checking 50+ bar opinions manually -- is exactly the kind of busywork AI was supposed to eliminate.

The Three Camps: Where Every State Falls

States break into three groups right now. Regulators like California, Florida, and New York have issued formal ethics opinions or proposed rule changes specifically addressing generative AI. Followers like Texas, Illinois, and Colorado have guidance that references AI indirectly through existing competence and confidentiality rules. Silent states -- roughly 15 as of early 2026 -- haven't addressed AI at all, which doesn't mean you're safe. It means you're operating without a net. The ABA's Model Rules still apply, and Rule 1.1 Comment 8 on technology competence is doing heavy lifting in jurisdictions that haven't spoken up.

Disclosure Requirements: The Biggest Split

This is where it gets messy. California's Proposed Formal Opinion 2024-1 says you must inform clients when AI is used in substantive legal work. Florida Bar Ethics Opinion 24-1 requires disclosure when AI materially affects representation. New York's guidance focuses on court-facing disclosure. But most states? They're silent on client disclosure entirely. The practical move: disclose anyway. No lawyer has ever been sanctioned for telling a client they used AI. Plenty have been sanctioned for not disclosing things the bar later decided they should have.

Confidentiality: The Universal Rule Nobody Follows Well

Every single state agrees on one thing -- you can't dump client data into AI tools without safeguards. Rule 1.6 confidentiality applies to AI inputs. That means understanding data retention policies, whether your prompts train the model, and where data is stored. Enterprise tiers from Anthropic and OpenAI don't train on your inputs. Free ChatGPT does. The number of lawyers still using free-tier tools with client information is genuinely alarming. If your state hasn't issued AI-specific guidance, default to the strictest confidentiality interpretation you can find.

Supervision and Responsibility: You Can't Blame the Machine

Every state that's addressed AI agrees: the lawyer is responsible for the output. You can't delegate judgment to Claude or ChatGPT and call it a day. ABA Formal Opinion 512 (July 2024) made this explicit -- generative AI is a tool, and the ethical obligations of competence, diligence, and supervision apply in full. For managing partners, this means having a firm-wide AI policy isn't optional anymore. Associates using AI without guardrails create liability for the entire partnership. The firms getting this right have written policies, approved tool lists, and mandatory output verification workflows.

How to Use This Map Practically

Don't just check your home state. Check every jurisdiction where you have active matters. A lawyer admitted in New York handling a California case needs to comply with both states' ethics frameworks. Bookmark our individual state pages -- we've built detailed guides for all 50 states plus DC, covering the specific ethics opinions, proposed rules, and practical compliance steps for each jurisdiction. When a new opinion drops, we update the relevant state page within 48 hours. This master guide links to every one of them so you always have a single starting point.

The Bottom Line: State-by-state AI ethics compliance isn't a one-time project. It's an ongoing obligation that changes every time a bar association issues new guidance. The firms that treat this as infrastructure -- not a checkbox -- are the ones that won't get caught flat-footed when the next round of formal opinions drops.