Oklahoma: AI Ethics Guidance for Law Firms

Bottom line for an Oklahoma firm: No OBA formal ethics opinion governs AI use yet. In state court, the Oklahoma Supreme Court’s March 2026 order (2026 OK 19) requires no AI disclosure in legal filings but makes every filer responsible for accuracy, and its May 2026 Reeves decision treats knowingly filing unverified AI citations as a knowing false statement. Two federal judges’ AI rules and one appellate court rule create verification obligations in Oklahoma proceedings today, and they differ on disclosure. Before Chief Judge Palk (W.D. Okla.), a filer who used generative AI discloses it, identifies the tool, and certifies that the AI-drafted portions were checked. Before Magistrate Judge Robertson (E.D. Okla.), no disclosure is required, but a reasonable inquiry requires human verification of cited authority. Before the Oklahoma Court of Criminal Appeals, Rule 1.17 requires that any AI-produced portion of a filing be verified, with no disclosure. The Mattox sanctions order is the live demonstration that courts will impose five-figure monetary penalties when verification fails.

Start with three artifacts: a written AI use policy, a citation and authority verification checklist for every federal and OCCA filing, and a court-by-court AI order log for the venues the firm appears in.

OBA Practice Management Guidance on AI Use

Citation: Oklahoma Bar Association, Management Assistance Program and Ethics Counsel, informal tip sheet on responsible AI use; Julie Bays, “Don’t Let AI Override Your Judgment” (OBA Practice Management). Companion coverage in the December 2025 Oklahoma Bar Journal and “It Is About Trust”.

Status: Non-binding informal guidance. The OBA Ethics Committee has not issued a formal opinion on AI use in legal practice. Monitor the OBA Ethics page for any future formal opinion.

What it requires

• Under Oklahoma RPC 1.1 Comment 6 (technology competence amendment adopted September 19, 2016), attorneys must keep abreast of changes in practice including the benefits and risks of relevant technology.
• Attorneys must verify AI-generated case citations against primary sources before filing. The guidance specifically warns that fabricated citations have been produced by general-purpose AI and by legal research platforms including Lexis, CoCounsel, and Vincent.
• Under Oklahoma RPC 1.6, attorneys must not upload client-identifying or matter-specific information to public AI tools without client consent.

What it recommends

• Ask AI for counterarguments and weaknesses in every output to counter sycophancy bias. See “Cross-Examining Your AI”.
• Cross-check all AI outputs against independent sources and maintain audit logs of prompts and outputs.
• Confirm court-specific AI affidavit or disclosure requirements before any filing. See OBA AI conference takeaways.

Notable gaps

The OBA has issued no formal ethics opinion interpreting any Oklahoma RPC rule in the AI context, no statewide mandatory AI disclosure rule covering all Oklahoma court filings, and no AI-specific CLE requirement.

What federal courts in Oklahoma require for AI use in filings

E.D. Okla. (Magistrate Judge Jason A. Robertson): Judge Robertson’s Candor, Accuracy, and Responsibility in Court Filings (updated January 6, 2026; tracker entry) covers filings in his criminal and civil cases “whether submitted by counsel or by a party appearing pro se.” It states that “This Court does not regulate, restrict, or require disclosure of the tools or assistance used to prepare filings,” but that “A reasonable inquiry requires human verification of cited authority,” and presenting a filing certifies that “The authorities cited exist and are accurately represented.” It replaced his September 2023 disclosure-and-certification order, which required filers who used generative AI to disclose it, identify the tool, and certify that the AI-drafted portions were checked. That order was in force when he sanctioned counsel in Mattox v. Product Innovations Research USA (E.D. Okla. Oct. 22, 2025; tracker entry).

W.D. Okla. (Chief Judge Scott L. Palk): Judge Palk’s AI Guidelines (linked from his chambers rules page) require parties that used generative AI to disclose that AI was used and identify the specific tool. The W.D. Okla. Bankruptcy Court’s General Order 23-01 (effective September 1, 2023) requires an attestation identifying the AI program and the portions it drafted, certifying that the document was checked for accuracy, and certifying that the program’s use did not disclose confidential information to any unauthorized party. See the tracker entry for Judge Palk’s order.

N.D. Okla.: No district-wide AI standing order has been confirmed. Before filing, firms should review the assigned judge’s chambers page at the N.D. Okla. site. The operative federal default is Fed. R. Civ. P. 11.

What Oklahoma state courts require for AI use in filings

Oklahoma Court of Criminal Appeals, Rule 1.17 Use of Generative Artificial Intelligence (adopted by order of February 18, 2026, 2026 OK CR 7, effective that day; tracker entry): When generative AI has been used in drafting any document for filing in the Court, the party or counsel must ensure that any AI-produced or AI-modified portion has been verified as accurate by a person responsible for the document. Generative AI is defined as “any type of artificial intelligence that generates content or data in response to a prompt or query by a user.” Failure to comply may result in sanctions including waiver of the affected issue(s) on appeal, striking of the non-compliant document from the record, and/or a finding of contempt. The verification duty is set out in a standalone rule of court rather than left to the candor duty under Oklahoma RPC 3.3.

Oklahoma Supreme Court, Re: Generative Artificial Intelligence, 2026 OK 19 (Mar. 23, 2026; docket SCAD-2026-17; tracker entry): “The Supreme Court does not discourage the use of AI in Oklahoma courts, provided its use complies with legal and ethical standards. Disclosure of AI assistance shall not be required in a legal filing. All persons filing documents in Oklahoma courts are responsible for the content and accuracy of the documents.” The order forbids no AI use and requires no disclosure in legal filings, but it places responsibility for accuracy on every filer. It does not say whether it displaces an individual judge’s case-specific or standing-order disclosure requirement, so firms should still check the assigned judge’s orders.

State ex rel. Oklahoma Bar Ass’n v. Reeves, 2026 OK 37 (May 27, 2026): In reciprocal discipline arising from a federal sanctions order in Alabama, the Supreme Court publicly reprimanded an Oklahoma-admitted lawyer who had added five ChatGPT-generated citations to federal filings without verifying them. Quoting its March 2026 order, the Court held: “If a lawyer knowingly uses AI to generate citations or analysis for a pleading and knowingly does not verify the accuracy before including it in the pleading, then the lawyer has knowingly made a false statement.” It also wrote that “There is nothing inherently problematic with the use of generative artificial intelligence in preparing legal materials, so long as attorneys abide by their duty to protect client confidentiality and recall their sacred duty to verify the pleadings they sign.” For Oklahoma lawyers, the practical effect is that lack of actual knowledge that a citation is fabricated may not be a defense under ORPC 3.3 when the lawyer knowingly skipped verification.

Mattox v. Product Innovations Research USA (E.D. Okla. 2025)

Magistrate Judge Jason A. Robertson, E.D. Okla., October 22, 2025. Plaintiffs’ counsel submitted eleven pleadings across summary-judgment briefing, motions in limine, and a sanctions motion. The pleadings contained fabricated case citations, erroneous citations, quotations of nonexistent law, and misstatements of law. The court found 28 false or misleading citations and violations of Fed. R. Civ. P. 11(b). Opposing counsel raised the alarm, and plaintiffs’ counsel initially characterized the problem as clerical and formatting errors. Judge Robertson rejected that framing, stating “the problem was not form, it was falsity.” The court publicly reprimanded counsel and imposed individual fines of $3,000, $2,000, and $1,000 ($6,000 in total), barred local counsel from sponsoring pro hac vice attorneys for twelve months, struck the affected filings, and ordered two firms to reimburse defense counsel’s $23,495.90 in fees and costs; it described the sanctions as “not punitive but restorative.” The order closes: “Before this Court, artificial intelligence is optional. Actual intelligence is mandatory.” Order PDF.

What AI-related rules are pending in Oklahoma?

No Oklahoma AI legislation targeted at the legal profession has been enacted. HB 3453 (2024) and HB 1916 / HB 1917 (2025) all died without enactment. Multiple 2026 session bills have been filed (HB 3545, HB 3546, SB 746) but none address the legal profession directly, and none have been confirmed enacted. The Oklahoma AG has signed a letter critical of federal preemption of state AI regulation, indicating continued state-level activity.

How do malpractice carriers in Oklahoma treat AI use?

Oklahoma has no state-specific lawyers’ professional liability carrier. Oklahoma attorneys typically carry coverage through national programs such as ALPS, CNA, and Lawyers Mutual. No Oklahoma-specific AI underwriting requirements have been confirmed from any carrier as of 2026-04-23. Carrier publications document three national exposure scenarios. First, hallucinated citations (likely covered as errors and omissions, but sanctions under Mattox sit outside any malpractice claim). Second, confidentiality breach via public AI tools (may trigger intentional-act or breach-of-confidentiality exclusions). Third, AI-generated content delivered directly to clients without attorney review (potential UPL exposure that LPL policies typically exclude). See ALPS analysis and ABA Journal coverage.

What does my Oklahoma malpractice carrier ask about AI at renewal?

No Oklahoma carrier has published Oklahoma-specific AI application items in materials reviewed for this entry. The substantive standard the carrier or broker will reference is Oklahoma RPC 1.1 Comment 6, RPC 1.6, and RPC 3.3. The Mattox sanctions record serves as the factual baseline for what verification looks like in practice. Plan on producing five artifacts: a written AI use policy; a citation and authority verification checklist with per-filing sign-off; a court-by-court AI order log covering E.D. Okla., W.D. Okla., and OCCA; vendor due-diligence records; and a supervision record for AI-assisted work product. Confirm application items directly with the broker; most LPL applications reviewed for this site do not yet ask about AI, but broker commentary reports LPL underwriters focusing on firms’ AI controls.

What documentation should a Oklahoma firm keep on file?

Month one (foundational)

1. (Owner: managing partner + firm administrator) Written AI use policy listing approved tools, categories of client information that may not be input into any tool (mapping to RPC 1.6), and required steps before any AI output is filed or transmitted.
2. (Owner: litigation lead) Citation and authority verification checklist with attorney sign-off on every federal and OCCA filing confirming each cited case, statute, and regulation has been verified against a primary source. The Mattox order is the factual baseline.
3. (Owner: litigation lead) Court-by-court AI order log covering at minimum E.D. Okla. (Judge Robertson: human verification of cited authority, no disclosure), W.D. Okla. (Judge Palk: disclose, identify the tool, and certify), and OCCA Rule 1.17 (verify all AI content). Treat N.D. Okla. as unconfirmed until checked against current chambers pages.

Months two and three (operational documentation)

1. (Owner: firm administrator + outside IT) Vendor data handling assessment for each AI tool: input retention for training, encryption, and whether vendor terms satisfy RPC 1.6.
2. (Owner: managing partner) Supervision protocol documenting who reviewed each AI-assisted draft, what changes were made, and who signed off before filing or transmittal, mapping to RPC 5.1 and 5.3.
3. (Owner: firm administrator) Training records for attorneys and non-lawyer staff covering firm AI policy and citation verification expectations.

Months four to six (per-matter discipline)

These are recurring practices, not one-time projects: each new matter exercises them.

1. (Owner: matter lead attorney) Client engagement letter AI clause and confidentiality consent disclosing whether AI tools may be used in the matter and what client information will and will not be shared.
2. (Owner: matter lead attorney) Incident response procedure if an AI hallucination reaches a filing: duty to notify the tribunal under RPC 3.3(a)(1), corrective filing, and preservation of the AI interaction record. Mattox shows that post-hoc characterization as a formatting error will be rejected.

We are building practitioner resources for firms working through this list. The newsletter covers new resources as they ship.