AI guidance across US adjudicators.

How to read this tracker. #

Every row in every table on this page carries (1) a current status, (2) the citation or formal name of the guidance instrument, (3) the date it was issued or, for negative entries, the date we last verified the absence of guidance, and (4) a link to the primary source. The link is always authoritative. If the surrounding text on this page is out of date, follow the link.

The status badges:

• ISSUED — formal opinion, rule, standing order, or policy has been published.
• TASK FORCE — the bar / court / tribunal has formed a task force or working group, but no formal guidance has issued yet.
• INTERNAL — for administrative tribunals: internal-only guidance exists (memos to ALJs, etc.) but no published rule for representatives.
• PROPOSED — rule was proposed but not adopted (typically withdrawn after comment, or pending).
• NO POLICY / NO ORDER / NO FORMAL OP. — we verified that no formal guidance exists as of the date stamp.

This page covers guidance directed at attorneys and representatives, not substantive AI rules directed at primary actors (consumers, employers, registrants). EEOC's AI guidance for employer hiring use of AI, for example, is not on this page; it is a different body of law.

Sources used and methodology. #

This tracker draws from primary sources where available (state bar websites, court orders, agency policy memoranda) and from comprehensive secondary trackers (Bloomberg Law, Law360, Justia 50-state survey, Reed Smith, Goodwin, Wilson Sonsini, ABA, Stanford CodeX RegLab) for confirmation. Where primary and secondary sources conflict, we follow the primary source. Where a secondary tracker reports an opinion we cannot find in the primary source, we mark the row as task force / unconfirmed pending verification.

What is NOT on this page:

• Substantive AI law affecting primary actors (consumer protection, employment discrimination, securities). This page is about practitioner conduct rules.
• Foreign jurisdictions. US federal + state + DC + admin only.
• Continuing legal education resources (CLE catalogues from each state). The bar opinion is the rule; CLE is implementation.
• Vendor-side guidance from AI tool providers. That belongs in the vendor-diligence playbook, not the regulatory tracker.

State bar AI ethics opinions — all 50 states + DC. #

Of 51 jurisdictions (50 states + DC), 21 have issued formal AI ethics guidance, 10 have active task forces or working groups, and 20 have no formal AI-specific guidance as of May 2026. ABA Formal Opinion 512 (July 2024) is treated as persuasive authority in most no-formal-opinion states.

Cross-reference: see A.06 — State Bar AI Opinions: A Comparative Tracker for narrative analysis of the leading opinions and the cross-state pattern. This table is the comprehensive index.

Last verified: May 9, 2026. If you find an error, email [email protected].

State court rules and standing orders on AI. #

State courts are slower than state bars to formally regulate practitioner AI use. Where state courts have issued rules, they typically take one of three forms: (a) state-supreme-court-level rule amendments parallel to bar opinions (often adopted after the bar opinion as the operationalization), (b) court-administrative-orders on disclosure and verification (rarer), or (c) individual-judge standing orders in trial courts.

As of May 2026, the cleanest published source for state-court-level AI rules is the Bloomberg Law tracker and the Law360 AI tracker. We do not duplicate their corpus here; we link out for the comprehensive view and call out structural patterns.

States with notable state-court-level AI activity (verified May 2026):

• California — Rule of Court 10.430 requires courts to either ban AI use by judicial officers / research attorneys or adopt a formal AI policy with verification and disclosure requirements.
• Florida — Miami-Dade and Palm Beach County administrative orders require attorneys and self-represented litigants to disclose generative AI use in court filings.
• New York — Multiple individual judges in the Commercial Division have issued AI standing orders; no state-court-wide rule yet.
• Texas — State court AI guidance limited to the bar's TRAIL initiative; individual judges in major counties have begun issuing orders.

Last verified: May 9, 2026.

Federal courts — circuits and districts. #

The federal-court layer is the noisiest. Since Judge Brantley Starr (N.D. Tex.) issued the first AI standing order in May 2023, hundreds of individual federal judges have issued their own orders. Court-wide standing orders are rare; individual-judge orders are the norm. The two canonical comprehensive trackers are Bloomberg Law and Law360; both are paywalled.

Circuit-wide guidance (as of May 2026)

Notable individual federal-judge / district-wide orders (representative)

Not exhaustive. Maintained as a representative cross-section of the most-cited orders. The full corpus is in Bloomberg / Law360.

Last verified: May 9, 2026.

Administrative tribunals — EOIR, SSA, VA, PTAB, TTAB, and the rest. #

Administrative tribunals are 18 months behind Article III courts on AI guidance. The two with formal published policies are EOIR (immigration courts, August 2025) and USPTO (PTAB / TTAB, April 2024). The rest — SSA / OHO disability, VA / BVA, NLRB, EEOC, SEC ALJs, FCC ALJs, FTC, FERC, MSPB, FLRA, Tax Court — have no published practitioner AI policy as of May 2026.

This is the layer practitioners are most exposed on. SSA disability hearings alone exceed the volume of all state and federal civil dockets combined; immigration-court hearings exceed all federal criminal dockets. Both have heavy AI use by counsel and limited regulatory guidance.

Last verified: May 9, 2026.

Cross-jurisdictional patterns. #

Reading across all the layers above, four patterns are visible:

1. Verification is universal. Every adopted guidance instrument requires attorneys to verify AI-generated authority before submission. Whether expressed as a Rule 11 verification duty (federal courts), a Rule 1.1 / 5.3 competence-and-supervision duty (state bars), or a sui-generis certification (federal-judge standing orders), the bottom line is the same: the lawyer is responsible for the words the AI produced.
2. Pre-emptive disclosure is the minority approach. Most guidance does not require pre-emptive disclosure of AI use. Florida Op. 24-1 says disclose where material to billing or costs. Most other state bars don't require disclosure absent client request. Federal-judge standing orders that require certification of AI use or non-use are the most aggressive disclosure regime, and they remain the exception rather than the rule.
3. Confidentiality is the high-stakes layer. Every adopted guidance instrument warns about uploading client information to public AI tools. The post-Heppner federal case law has begun to give that warning teeth: Heppner held that public AI submission can defeat privilege. That makes the confidentiality directive in the bar opinions enforceable in collateral litigation, which is a sharp escalation.
4. The admin-tribunal gap. The volume-heaviest adjudicative bodies (SSA, immigration, VA) have the least guidance. This is exactly inverted from where practitioner exposure actually lives. Disability appeals and immigration filings are where AI hallucinations show up most often per filing volume, and where formal regulatory infrastructure is thinnest.

If your matter spans multiple jurisdictions. #

Many matters cross divides. A federal civil case in S.D.N.Y. with a state-court appeal to the New York Appellate Division and a related employment matter at EEOC can implicate three different rule regimes. Practical sequence:

1. Start with the most restrictive rule that applies. If any tribunal in your matter has a verification-with-certification standing order, comply with it for all filings.
2. For confidentiality, default to closed-loop AI tools regardless of forum. The strictest privilege regime in any jurisdiction touching the matter governs your tooling choice.
3. For disclosure, comply with the strictest forum-specific rule but do not pre-emptively disclose AI use to forums that do not require it (some clients view broad disclosure as waiving privilege; conserve it for where required).
4. Document the AI-use decision tree for the matter file. If a §10(a)(4) challenge or a sanctions motion later turns on what AI use you made, the contemporaneous documentation is the evidence you have.

For arbitration overlays specifically, see A.11 — AI in US Arbitration and A.12 — AI in Arbitration: Practice Playbook for Firms.

Citations and further reading. #

Comprehensive secondary trackers:

• Justia — AI and Attorney Ethics Rules: 50-State Survey (free).
• Steno — Legal AI Rules by State (free).
• Responsible AI in Legal Services (RAILS) — AI Orders Resource.
• Bloomberg Law — Federal Court Judicial Standing Orders on AI (paywalled).
• Law360 — Tracking Federal Judge Orders on AI (paywalled).
• Bloomberg Law — State Legal Ethics Guidance on AI (paywalled).

ABA-level guidance:

• ABA Formal Opinion 512 (July 2024) — the federal anchor for state bar opinions.

IXSOR cross-references:

• A.06 — State Bar AI Opinions: A Comparative Tracker — narrative analysis of leading state opinions.
• A.07 — ABA Formal Opinion 512: An Implementation Playbook.
• A.05 — NC State Bar 2024 FEO 1.
• A.11 — AI in US Arbitration.
• A.12 — AI in Arbitration: Practice Playbook for Firms.

This page is a comprehensive tracker of formal guidance directed at attorneys and representatives on AI use. It is not legal advice, does not establish an attorney-client relationship, and does not predict how any specific bar, court, or tribunal will rule on facts not before it. Per-row "as of" dates document last verification. Primary-source links are authoritative; if a row text disagrees with the linked source, the linked source controls.

About the author.