Lawyer Use of AI in 2026: Bar Rules, Court Filings, and What You Must Certify
· LawDiver
A practitioner's map of ABA Formal Opinion 512, state bar AI ethics opinions, statewide court rules, federal local rules, and judge standing orders — with enactment and effective dates and a link to every authority.
Lawyer Use of AI in 2026: Bar Rules, Court Filings, and What You Must Certify
Current as of September 9, 2026. This article is general information for lawyers and legal technologists, not legal advice. Bar opinions and court orders change quickly; confirm every primary source before you rely on it in a live matter.
There is still no nationwide ban on lawyers using generative AI. There is also no nationwide free pass. What exists instead is a layered stack:
- Existing Rules of Professional Conduct (competence, confidentiality, communication, candor, fees, supervision), read onto AI by ABA Formal Opinion 512 (July 29, 2024).
- State-bar ethics opinions and practical guidance that apply those rules to generative AI — mostly advisory, a few with unusual force.
- Statewide court rules and judicial-branch policies that bind filings in that court system.
- Federal Rules of Civil Procedure (especially Rule 11), plus district local rules and judge-specific standing orders that can require disclosure or certification even when your state bar does not.
The practical failure mode is always the same: a lawyer treats the bar opinion as the whole answer, then files before a judge whose standing order demands a certificate the bar never mentioned. Check the judge, then the court, then the bar — in that order — before the brief goes out.
What every lawyer is already under
ABA Formal Opinion 512 (Generative Artificial Intelligence Tools), issued July 29, 2024 by the ABA Standing Committee on Ethics and Professional Responsibility, is the national baseline. It does not create new Model Rules. It maps existing rules onto generative AI. ABA formal opinions are persuasive, not binding, until a state adopts them or a court relies on them — but nearly every later state opinion tracks this framework.
| Duty | Model Rule | What Formal Op. 512 emphasizes |
|---|---|---|
| Competence | 1.1 (incl. tech competence comment) | Understand the tool's capabilities and limits; stay current as products change |
| Confidentiality | 1.6 | Evaluate data retention, training, and vendor terms before putting client information into a tool |
| Communication | 1.4 | Disclose AI use when it is material to the representation or the fee |
| Candor to the tribunal | 3.3 | Independently verify analysis, quotations, and citations before filing |
| Reasonable fees | 1.5 | Do not bill for time you did not spend; tool costs only with a proper expense arrangement |
| Supervision | 5.1 / 5.3 | Own AI output the way you own a junior associate's or vendor's work |
ABA Formal Opinion 512 does not require lawyers to disclose generative-AI use to every tribunal as a matter of ethics. Disclosure to courts is driven by court rules and standing orders, not by the ABA opinion alone.
Secondary write-up: ABA news announcement (July 29, 2024).
Authority hierarchy (read this before the tables)
Documents in this area do not carry equal weight. From least to most binding for a filing:
- ABA formal opinion — persuasive only.
- State-bar ethics opinion / practical guidance — usually advisory; still what a disciplinary panel will measure you against.
- Statewide court rule or judicial-branch policy — binds practice in that court system (for example, Florida AOSC26-12 / Rule 2.515(d)(2), 22 NYCRR Part 161, Illinois Supreme Court AI Policy).
- District local rule — binds every judge in that district (for example, N.D. Tex. Local Civil Rule 7.2(f)).
- Judge standing order / chambers rule — binds that courtroom only, and can be stricter than anything above it.
If your state bar says "no general duty to disclose AI" and your assigned judge requires a certificate, the certificate controls the filing.
State bar ethics opinions and formal guidance
Numbered or formal ethics opinions (and closely related Board-approved guidance) that specifically address lawyer use of generative AI. Dates are face, adoption, or effective dates as stated on the primary instrument.
| Jurisdiction | Authority | Source |
|---|---|---|
| ABA (national) | Formal Opinion 512 (issued July 29, 2024). Persuasive. Competence, confidentiality, communication, candor, fees, supervision. | FO 512 PDF; ABA announcement |
| Alabama | Formal Opinion 2026-01 (2026). Advisory. Existing RPC apply; verify outputs; discusses Mata v. Avianca and recent Alabama sanctions examples. | FO 2026-01 PDF |
| Alaska | Ethics Opinion 2025-1 (April 23, 2025). Advisory. Competence, confidentiality, supervision, candor, verification. | Op. 2025-1 PDF |
| California | COPRAC Practical Guidance for the Use of Generative AI (Nov. 16, 2023; Board-approved update May 14, 2026). Board-approved guidance (not a numbered ethics opinion); proposed RPC AI amendments pending. Competence (incl. agentic AI), confidentiality, verification, fees, supervision, communication. | Practical Guidance PDF; Ethics & Technology Resources; proposed RPC amendments |
| District of Columbia | Ethics Opinion 388 (April 2024). Advisory. Competence, confidentiality, candor, fees, supervision; preserve AI interactions in the client file. | DC Bar Op. 388 |
| Florida | Ethics Opinion 24-1 (January 19, 2024). Advisory. Confidentiality and consent for third-party tools; supervise/verify; reasonable fees; advertising/chatbot disclaimers. | Op. 24-1 PDF |
| Kentucky | KBA E-457 (March 15, 2024). Advisory. Tech competence; verify authorities; keep client-identifying data out of public tools; adjust fees if AI cuts time. | KBA E-457 PDF |
| Mississippi | Ethics Opinion No. 267 (November 14, 2024). Advisory. Confidentiality, verification, fees; informed consent in certain Rule 1.4 situations. | MS Bar Op. 267 |
| New Mexico | Formal Ethics Advisory Opinion 2024-004 (September 24, 2024). Advisory. Competence, confidentiality, verification, conflicts, supervision, communication, fees. | Op. 2024-004 PDF |
| New York City Bar | Formal Opinion 2024-5 (August 7, 2024). Advisory. Maps competence, confidentiality, supervision, and candor onto generative AI; routine embedded AI generally not a disclosure trigger. | NYC Bar Formal Op. 2024-5 |
| New York City Bar | Formal Opinion 2025-6 (December 22, 2025). Advisory. Consent, confidentiality, and verification when AI records, transcribes, or summarizes client conversations. | NYC Bar AI ethics hub |
| North Carolina | 2024 Formal Ethics Opinion 1 (adopted November 1, 2024). Authoritative in NC. Competence; vendor security diligence; supervise; verify before filing; bill only for time spent. | 2024 FEO 1 |
| Oregon | Formal Opinion 2025-205 (Board approved February 2025). Advisory. May use AI subject to competence, confidentiality/consent for open tools, supervision, candor, billing; verify citations. | OSB Op. 2025-205 PDF |
| Pennsylvania / Philadelphia | Joint Formal Opinion 2024-200 (May 2024). Advisory. Understand the technology; verify citations; protect confidences; informed consent where use is material. | Joint Formal Op. 2024-200 PDF |
| Texas | Professional Ethics Committee Opinion 705 (February 2025). Advisory. Competence; confidentiality (Tex. Rule 1.05); independently verify all AI output; do not bill for unworked time. | Opinion 705 PDF; TCLE Opinion 705 |
| Virginia | Legal Ethics Opinion 1901 (VSB Council June 12, 2025; Supreme Court of Virginia approval November 24, 2025). Approved by the Supreme Court of Virginia (unusual force). Reasonable fees and generative AI: a lawyer need not cut a reasonable value-based fee solely because AI saved time. | LEO 1901 PDF |
| Washington | WSBA Advisory Opinion 2025-05 (2025). Advisory (not a WSBA Board "position" in the formal sense). Applies competence, diligence, confidentiality, communication, candor, supervision, and billing to AI-enabled tools (including agentic AI). | WSBA |
| West Virginia | Lawyer Disciplinary Board L.E.O. 24-01 (June 14, 2024). Formal advisory. Among the strictest client-consent positions: informed consent to generative AI use, confirmed in writing; verify; supervise; review vendor privacy. | LEO 24-01 PDF; WV ODC LEO index |
Related official guidance (not numbered formal opinions)
These are still primary materials lawyers cite in practice. They are weaker than numbered opinions or court rules, but they are official.
| Jurisdiction | Authority | Source |
|---|---|---|
| New York (statewide bar) | NYSBA Task Force on AI — Report and Recommendations (April 6, 2024). Task-force report, not a numbered ethics opinion. | NYSBA Task Force report PDF |
| New Jersey | Supreme Court Preliminary Guidelines on NJ Lawyers' Use of AI (notice January 24, 2024). No new duties; existing rules govern; no general filing-disclosure duty. | NJ Courts notice |
| Missouri | Informal Opinion 2024-11 (April 25, 2024). Competence, confidentiality, verification, independence, supervision, fees. | Informal Op. 2024-11 |
| Georgia | State Bar Generative AI Toolkit (last updated February 20, 2026). Explicitly not a formal advisory opinion. | gabar.org |
| Ohio | Board of Professional Conduct AI for Lawyers and Judicial Officers ethics guide (published April 2026; widely reported June 2026). Non-binding staff guidance; not a Board advisory opinion. | bpc.ohio.gov |
| Illinois ARDC | The Illinois Attorney's Guide to Implementing AI (October 24, 2025). Complements the Illinois Supreme Court AI policy. | Illinois Courts AI policy announcement |
| Michigan | Artificial Intelligence for Attorneys: FAQ (November 18, 2024; fees Qs added February 11, 2025). FAQ guidance; judicial ethics opinion JI-155 (October 27, 2023) addresses judges. | michbar.org |
| Arizona | Steering Committee on AI and the Courts — Generative AI Ethical Best Practices (November 14, 2024). Best practices for lawyers and judges. | Arizona courts / committee materials via state judiciary |
States without a dedicated generative-AI ethics opinion
Absence of a numbered AI opinion does not mean AI is unregulated. Those jurisdictions still enforce competence, confidentiality, candor, fees, and supervision under existing rules, and ABA Formal Opinion 512 is the usual persuasive framework. Treat "no AI opinion found" as a research status, not a green light to skip verification.
Statewide court rules and judicial-branch policies
This is where filing practice actually changes statewide — independent of bar opinions.
| Court system | Authority | Source |
|---|---|---|
| Florida | Fla. R. Gen. Prac. & Jud. Admin. 2.515(d)(2) via AOSC26-12. Opinion May 28, 2026; amendments effective June 15, 2026. Signer represents that legal authorities identified in the filing exist and are accurately cited; sanctions after notice and opportunity to be heard. Statewide replacement for varied circuit AI disclosure/certification AOs; circuits may not impose local AI disclosure/cert requirements. | AOSC26-12 PDF |
| New York UCS | 22 NYCRR Part 161 (Use of Artificial Intelligence Technology). Added by Admin. Order March 25, 2026; effective June 1, 2026. AI use permitted; no system-wide duty to disclose AI use; filers must independently ensure no fabricated authorities; individual courts may adopt Appendix A model certification rule. | NYSBA summary of Part 161 |
| Illinois | Illinois Supreme Court Policy on Artificial Intelligence. Announced Dec. 18, 2024; effective January 1, 2025. Permits AI under existing RPC / Code of Judicial Conduct; users accountable for final work product; thorough review required; no AI disclosure required in pleadings under the policy. | Announcement; Policy PDF |
| California judiciary | Cal. Rules of Court 10.430 (+ Standard 10.80). Effective September 1, 2025 (court policies due Dec. 15, 2025). Courts that permit generative AI must adopt a written use policy covering privacy, bias, human review, and disclosure when content is entirely AI-generated (court-system facing framework). | Rule 10.430; Rules PDF (eff. Sept. 1, 2025) |
| Arkansas | Administrative Order No. 25 (AI) + ARPC amendments. Adopted Dec. 11, 2025 (2025 Ark. 220 / 221); effective immediately. Court-system AI order plus professional-conduct amendments emphasizing accountability/supervision. | 2025 Ark. 117 (Justia context) |
| Colorado | Rule Change 2026(02) — Colo. RPC preamble/scope and Rule 1.1 comments. January 8, 2026. AI-specific comment amendments to the Rules of Professional Conduct (competence / independent judgment). | Rule Change 2026(02) PDF |
| Rhode Island | In re Amendments to Article V, Rule 1.1 + Interim Generative AI Guidelines. June 17, 2026. Binding competence-comment amendment plus advisory interim guidelines. | RI Supreme Court order/guidelines PDF |
| Oklahoma (criminal appellate) | OCCA Rule 1.17 (Use of Generative AI). February 18, 2026 (2026 OK CR 7). Verification / accountability for filings in the Court of Criminal Appeals. | OCCA Rule 1.17 |
| Montana (local districts) | Fourth Judicial Dist. Rule 3.G (Oct. 31, 2025); Thirteenth Judicial Dist. Rule 35 (Jan. 1, 2026). Local disclosure + verification requirements for generative AI in pleadings (county/district scope only). | Montana Judicial Branch |
| Delaware | Supreme Court interim GenAI policy. Adopted Oct. 21, 2024. Primarily judges and court staff (not a general lawyer-filing disclosure rule). | courts.delaware.gov |
Florida's statewide filing representation (what you must say)
Florida's 2026 approach is the cleanest "statewide filing representation" model currently in force:
- By signing, the filer represents that cited legal authorities exist and are accurately cited.
- That representation applies to attorneys and unrepresented parties.
- Courts may sanction inconsistent filings after notice and an opportunity to be heard (reprimand, contempt, striking, dismissal, costs, fees, or other sanctions).
- AOSC26-12 expressly preempts circuit-level AI disclosure/certification administrative orders.
That is accuracy-of-authorities certification, not "I used ChatGPT" disclosure. The distinction matters for checklists.
New York Part 161 (what you must — and need not — say)
Under 22 NYCRR Part 161 (eff. June 1, 2026):
- AI tools are permitted for preparing court papers.
- There is no statewide duty to disclose AI use on the papers.
- Filers must carefully review and independently ensure the paper contains no fabricated or fictitious cases, statutes, or other material.
- Individual courts may adopt the Appendix A model rule under which a signature certifies the absence of fabricated AI content.
- Individual judges may still impose stricter part rules — check the assigned judge.
Federal rules, circuits, districts, and standing orders
Federal Rules of Civil Procedure (no AI-specific rule yet)
There is no Federal Rule of Civil Procedure that uses the words "artificial intelligence" or "generative AI" as of September 9, 2026.
What already binds every federal civil filing:
- Fed. R. Civ. P. 11(b) — by presenting a paper, a signer certifies (after inquiry reasonable under the circumstances) that legal contentions are warranted by existing law or a nonfrivolous argument for change, and that factual contentions have evidentiary support.
- Related duties under local rules and professional-conduct standards incorporate the same accuracy obligation.
A May 29, 2026 rules suggestion (26-CV-21) to the Judicial Conference Advisory Committee on Civil Rules proposes a Rule 11(a)-style certification that cited legal authorities exist. That is early-stage federal rulemaking, not an adopted amendment. Suggestion materials: uscourts.gov 26-CV-21 PDF.
Federal courts of appeals
| Court | Authority | Source |
|---|---|---|
| Fifth Circuit | Declined to adopt a court-wide special AI briefing rule (June 2024). Existing accuracy duties continue; no circuit-wide AI certificate. | Court decision on proposed rule PDF |
| Ninth Circuit | Circuit Advisory Committee Note to Ninth Circuit Rule 32-1 (signature attestation and generative AI). Dec. 2025 revisions; handbook materials effective June 1, 2026. Advisory note interpreting existing FRAP signature / attestation obligations. | Check current Ninth Circuit rules / advisory committee notes on ca9.uscourts.gov |
Appellate courts have generally preferred existing candor and signature rules over new AI-disclosure regimes. District practice is where most AI certificates live.
Federal district local rules (district-wide)
Local rules bind every judge in the district. Texas examples show the modern pattern:
| District | Authority | Source |
|---|---|---|
| N.D. Texas | Local Civil Rule 7.2(f) (Disclosure of Use of Generative Artificial Intelligence). Effective September 2, 2025. Brief prepared with generative AI must disclose that fact on the first page under the heading "Use of Generative Artificial Intelligence"; absence of the heading certifies that no part was prepared with generative AI. | Local Rule 7.2(f) tracker; txnd.uscourts.gov |
| E.D. Texas | Local Rule CV-11(g) (and related amendments). AI caution language in local rules; further standing orders in divisions. Pro se / technology-assisted filings remain the filer's responsibility for accuracy. | txed.uscourts.gov |
| D. Nebraska | En banc local-rule amendment (December 2024). All parties responsible for accuracy/reliability of briefing regardless of generative-AI assistance; points to Rule 11(b). | Check D. Neb. local rules on ned.uscourts.gov |
Judge standing orders (chambers rules)
The chambers-order wave began with Judge Brantley Starr (N.D. Tex.) on May 30, 2023, requiring a certificate that either no generative AI drafted any portion of filings, or that any AI-drafted language was checked for accuracy by a human using print reporters or traditional legal databases. Chambers page: Judge Starr (N.D. Tex.).
Since then, well over 100 federal and state judges have issued standing orders that fall into recurring templates:
- Disclose + certify verification — identify that AI was used (sometimes naming the tool and the portions) and certify human verification of citations and facts.
- Certify non-use or verified use — Starr-style binary certificate.
- Rule 11 reminder / caution — no separate certificate, but an express warning that Rule 11 applies to AI-assisted filings.
- Prohibit AI in filings — a minority approach (examples historically include some chambers in N.D. Ohio and S.D. Ohio).
Illustrative primary-source examples (not exhaustive):
| Judge / body | Authority | Source |
|---|---|---|
| Judge Brantley Starr (N.D. Tex.) | May 30, 2023. Certificate of non-use or human verification. | Chambers page |
| Judge James L. Graham (S.D. Ohio) | Oct. 21, 2025. "Disclosure of the Use of Generative AI" declaration identifying portions and platform; certify verification under Rule 11. | Standing order PDF |
| E.D. Tex. (e.g., JDL order) | April 9, 2025. Certificate disclosing use/non-use, tool, how used, and accuracy check. | JDL Standing Order on AI PDF |
Living trackers that link chambers PDFs (use to find your judge, then click through to the court PDF):
- Legal AI Governance — federal/state court orders tracker
- Firm client alerts that catalog standing orders (for example, Husch Blackwell overview)
Do not treat a tracker as the authority. Treat it as an index. The PDF on the court's site is the authority.
Bankruptcy courts, agencies, and specialty forums
Some bankruptcy judges and specialty tribunals have their own AI standing orders or memos (for example, EOIR policy memoranda on generative AI in immigration proceedings). Those are forum-specific. If you practice before a specialty tribunal, check that tribunal's clerk's office and judge pages the same way you check a district judge's individual practices.
What representations you typically must make in court filings
There is no single national form. In practice, representations fall into five buckets. Map your matter against all five.
1. The signature / Rule 11 (or state analogue) representation
Always present. Signing a federal pleading is already a certification under Rule 11(b). Many state systems have analogues (New York 22 NYCRR 130-1.1 frivolous-conduct rules; Florida Rule 2.515 representations; etc.). AI does not dilute that certification.
2. Accuracy-of-authorities certifications (statewide)
Example: Florida Rule 2.515(d)(2) (eff. June 15, 2026) — cited authorities exist and are accurately cited. This is the emerging "statewide" model that replaces patchwork local AI disclosure orders.
3. AI-use disclosure on the face of the brief
Example: N.D. Tex. Local Civil Rule 7.2(f) — first-page heading "Use of Generative Artificial Intelligence" when generative AI prepared the brief; silence equals certification of non-use.
4. Separate AI certificate / declaration
Common chambers template:
- State whether generative AI was used.
- Identify the tool (ChatGPT, Copilot, CoCounsel, Harvey, Gemini, etc.).
- Identify the portions prepared with AI assistance.
- Certify that a human reviewed source material and verified citations, quotations, and legal propositions.
- Acknowledge continuing Rule 11 responsibility.
5. Client-side representations (ethics, not captions)
Separate from court captions:
- Informed consent / disclosure to the client when AI use is material (ABA 512; Florida 24-1; Oregon 2025-205 for open models; West Virginia LEO 24-01 written consent).
- Engagement-letter and billing representations about fees, expenses, and efficiency.
- Advertising / chatbot disclaimers when AI interacts with prospective clients (Florida 24-1).
Practical filing checklist (use this before you hit send)
- Identify every forum: state trial court, state appellate court, federal district, bankruptcy, agency.
- Pull the assigned judge's individual practices / standing orders.
- Pull the district or circuit local rules for AI, technology, or Rule 11 add-ons.
- Pull any statewide court AI rule (Florida 2.515; New York Part 161; Illinois policy; etc.).
- Confirm your state's bar ethics opinion / guidance for confidentiality and client consent.
- Decide whether the filing needs: (a) no extra certificate, (b) first-page disclosure, (c) separate declaration, (d) accuracy-of-authorities representation already embedded in the signature rule.
- Independently verify every citation and quotation against a primary reporter or a trusted legal database — not against the model's own assurance that the cite is real.
- Keep a verification workpaper in the matter file (what was checked, where, by whom).
Independent verification is the one duty that appears in essentially every opinion and order. If you only institutionalize one habit, institutionalize that one.
CiteDiver produces a verified, audited trail for that habit: it extracts citations from a brief, resolves them against LawDiver's live caselaw corpus, scores existence and accuracy, and leaves a workpaper you can keep with the matter file for Rule 11 / Rule 3.3 / Florida 2.515-style accuracy representations. The verification duty remains the lawyer's; CiteDiver is the audit-trail tool.
How CiteDiver helps satisfy the verification duty
The lawyer still owns competence and candor. CiteDiver is built to document the check, not to replace it:
- Extracts citations from the draft and resolves them against LawDiver's live caselaw corpus (existence, reporter string, and accuracy signals).
- Leaves a dated workpaper you can file with the matter for Rule 11, Rule 3.3, or Florida 2.515-style accuracy representations.
- Fits the same workflow as a "trusted legal database" check — useful when a standing order demands human verification of AI-assisted cites.
For API-level cite checking in legal products, see the caselaw API for legal AI builders.
Sanctions reality check
The enforcement story is still mostly hallucinated authorities, not "you used AI without a sticker."
- Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023) — early high-profile sanctions for ChatGPT-invented cases.
- Park v. Kim, 91 F.4th 610 (2d Cir. 2024) — Second Circuit sanctions practice reinforcing that counsel remains responsible for fabricated or unchecked authorities in AI-assisted filings.
- Dozens of later federal and state sanctions / fee-shifting / referral orders for fabricated citations, fake quotations, and false statements about how a brief was prepared.
- Discipline can travel with the same facts under candor and misconduct rules even when the court uses Rule 11 rather than an AI-specific order.
"The AI hallucinated" is not a safe harbor. The lawyer signed the paper.
How to build one policy for a multistate practice
You do not need fifty AI policies. You need three artifacts:
- Firm AI-use policy written to ABA Formal Opinion 512 duties plus the strictest verification language in any jurisdiction you touch (California's proposed verify-every-output amendments are a useful ceiling even while pending).
- Vendor diligence checklist (training on inputs, retention, subprocessors, confidentiality commitments, audit rights) run before any tool is approved firm-wide.
- Per-filing court-order check on the matter checklist — the step most teams skip.
Write the policy to duties, not to a single state's opinion. Update it on a fixed cadence. When a new statewide rule lands (Florida 2.515; New York Part 161; Colorado RPC comments), patch the filing checklist first and the ethics memo second.
Frequently asked questions
May lawyers use generative AI in 2026?
Yes. No U.S. jurisdiction surveyed here bans lawyers from using generative AI as such. Use is conditioned on existing ethics duties and on any court-specific disclosure or certification rules that apply to the filing.
Is ABA Formal Opinion 512 binding?
No. It is persuasive national guidance issued July 29, 2024. It becomes controlling only when a state adopts it or a court relies on it. Most state opinions track it closely. Primary text: ABA Formal Opinion 512 PDF.
Do I have to tell the court I used AI?
It depends on the forum. Many courts require no disclosure (Illinois policy; New Jersey guidelines; New York Part 161 statewide floor). Others require first-page disclosure (N.D. Tex. Local Civil Rule 7.2(f)) or a separate declaration (many chambers orders). Florida's statewide rule focuses on accuracy of authorities rather than AI-use disclosure. Always check the assigned judge.
What is the most common requirement across bar opinions?
Independent verification of AI outputs — especially citations and quotations — before they reach a client or a tribunal. That duty flows from competence and candor, not from a special "AI rule." Cases like Mata v. Avianca and Park v. Kim show what happens when verification fails.
What must Florida filers represent after June 15, 2026?
Under amended Rule 2.515(d)(2), the signer represents that legal authorities identified in the filing exist and are accurately cited. Local circuit AI disclosure/certification administrative orders are preempted by AOSC26-12.
What does New York Part 161 change?
Effective June 1, 2026, the Unified Court System permits AI tools for court papers, declines a system-wide disclosure mandate, requires independent review to prevent fabricated material, and lets individual courts adopt an optional Appendix A certification model. See the NYSBA Part 161 summary.
Does West Virginia really require written client consent?
West Virginia Lawyer Disciplinary Board L.E.O. 24-01 (June 14, 2024) is widely treated as the strictest client-consent position: informed consent to generative AI use, confirmed in writing. Confirm the opinion text for your matter; it is advisory but heavily relied on in WV practice discussions. Index: WV ODC Legal Ethics Opinions.
Are there federal AI filing rules?
Not in the Federal Rules of Civil Procedure themselves as of September 9, 2026. Rule 11 already supplies the accuracy certification. District local rules and judge standing orders supply most AI-specific disclosure language. A 2026 rules suggestion (26-CV-21) would add an authorities-exist certification; it is not adopted law.
My state is not in the formal-opinion table. Is AI unregulated there?
No. Existing Rules of Professional Conduct still apply. Lack of a numbered AI opinion means you fall back to those rules plus ABA Formal Opinion 512 and any court orders in your forums.
How often should this guide be re-checked?
Before any filing in an unfamiliar court, and whenever your firm updates its AI policy. Statewide rules in 2025–2026 (Illinois, California Rule 10.430, Florida 2.515, New York Part 161, Colorado RPC comments, Rhode Island) show how quickly the court layer can move.
Related reading
- CiteDiver — citation verification and audit trail against LawDiver's live caselaw corpus
- Caselaw API for legal AI builders
Article current as of September 9, 2026.