District of Columbia: AI Ethics Guidance for Law Firms

Bottom line for a 5-50 attorney DC firm: D.C. Bar Legal Ethics Opinion 388 (April 2024) applies the D.C. Rules of Professional Conduct to generative AI, and D.C. Court of Appeals Order M284-24 (effective May 6, 2025) added technology to the Rule 1.1 competence comment and extended confidentiality and supervision duties to outside service providers. The D.C. Court of Appeals struck a brief containing fabricated citations and referred the lawyer to Disciplinary Counsel in Douglas v. Deutsche Bank (September 3, 2026). Neither the D.D.C. nor the D.C. Circuit has a court-wide AI rule, but Judge Walton’s general order for civil cases addresses generative AI, so the assigned judge’s orders are worth checking before any federal filing.

Start with: a written AI use policy, a vendor diligence record for each AI tool in use, and a citation verification step for any filing that relies on AI-assisted research.

D.C. Bar Legal Ethics Opinion 388

Citation: D.C. Bar Legal Ethics Opinion 388, “Attorneys’ Use of Generative Artificial Intelligence in Client Matters” (published April 2024), issued by the D.C. Bar Legal Ethics Committee. Primary source.

Status: Formal ethics opinion of the Legal Ethics Committee. It is advisory, not a binding rule. As of 2026-09-30, it remains the Committee’s only opinion on AI: the opinion index lists Opinions 389 through 392 (August 2025 to March 2026), none of which addresses AI.

What it says

• Competence (Rule 1.1): “Before using any particular form of GAI, attorneys should have a reasonable and current understanding of how it works and what it does,” with due regard for its potential dangers (including hallucinations), its limitations, and its cost.
• Confidentiality (Rule 1.6): Before using a tool in a client matter, a lawyer considers two questions: whether information provided to the tool will be visible to the provider or others outside the attorney-client relationship, and whether the lawyer’s interactions could affect answers given to later users in a way that reveals client information.
• Supervision (Rules 5.1, 5.3): Lawyers “should take reasonable measures” to ensure that supervised lawyers’ and nonlawyers’ use of generative AI conforms to the Rules. The opinion describes steps a firm “could consider,” such as requiring staff to confirm that confidentiality will be protected before using a tool, requiring verification of outputs, or limiting staff to a tool the firm has vetted.
• Candor (Rules 3.3, 3.4): Attorneys “should have a reasonable basis for trusting the GAI outputs, or must review and validate GAI outputs,” before incorporating them in work product or relying on them in a legal proceeding. The opinion uses Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023) as its cautionary example, and notes that Rule 3.4(c) applies when a tribunal has issued rules or orders governing generative AI.
• Fees (Rule 1.5): “Lawyers whose fee agreements provide for fees based solely on time spent may only bill for the time the lawyers actually spend, even if the GAI reduces the time the lawyers devote to the matter.” Lawyers may pass through out-of-pocket generative AI expenses where the client has agreed to pay out-of-pocket expenses.
• Client file (Rule 1.16(d)): “A lawyer should consider whether specific interactions with GAI in connection with a client matter should be retained as part of the client file.”

Diligence steps it describes

• Before adopting a tool, look to news coverage, disinterested third-party testing, and the experience of trusted peers, and compare the tool’s output with known work product.
• Where a tool’s terms raise confidentiality concerns, consider negotiating with the vendor for better confidentiality terms, “an upgrade to a paid product with better terms,” or a tool that offers “a zero data retention policy.”

Rules cited

Rules 1.1, 1.2, 1.5, 1.6, 1.16, 3.3, 3.4, 5.1, 5.3, 8.4.

Notable gaps

Opinion 388 does not address a general duty to tell clients that AI is being used on their matter; it discusses client consent only for sharing confidences with a tool and communicating expected AI costs under Rule 1.5(b). It does not address AI-generated expert reports, settlement communications, or transactional documents, and it does not distinguish AI tools embedded in research platforms from standalone generative AI. At the time it was written, the Court of Appeals was still considering the Rule 1.1 technology comment that Order M284-24 later adopted (the opinion’s footnote 9 describes that proposal).

D.C. Court of Appeals Order M284-24: RPC amendments

Citation: D.C. Court of Appeals Order No. M284-24, filed April 7, 2025. An April 24, 2025 order supplied the effective date: “The amendments will go into effect on May 6, 2025.”

Status: Binding amendments to the D.C. Rules of Professional Conduct and their comments, in force since May 6, 2025. None of the amended text mentions artificial intelligence.

What changed:

• Rule 1.1, Comment [5]: Competent handling now includes “use of methods, procedures, and technology meeting the standards of competent practitioners.”
• Rule 1.6(f): A lawyer must exercise reasonable care to prevent “the unauthorized access to confidences or secrets of a client,” in addition to disclosure or use by those whose services the lawyer uses. This bears on AI tools whose providers may access client data.
• Rule 1.6, Comment [40]: Addresses precautions when transmitting or storing client confidences, with reasonableness factors including the cost of security measures and the difficulty of implementing safeguards.
• Rule 5.3, Comments [3] and [4] (new): Address a lawyer’s reasonable efforts when using nonlawyer service providers, with examples including a document management company, a printing and scanning vendor, and “an internet-based service to store client information.” Applying these comments to AI tool vendors is this site’s reading; the comments do not name AI.
• Rule 4.4(c) (new): A lawyer who begins examining a writing and only then realizes it was inadvertently sent must stop examining it, notify the sender, and follow the sender’s instructions.

What federal courts in DC require for AI use in filings

The D.D.C. standing orders page lists no court-wide AI order, and the court’s Local Rules (September 2026 edition) contain no AI provision. The operative federal default is Fed. R. Civ. P. 11.

At the chambers level, Senior Judge Reggie B. Walton’s General Order for Civil Cases includes a paragraph titled “Use of Generative Artificial Intelligence.” It permits generative AI tools but expects counsel to comply with all legal and ethical obligations, including ensuring that representations about the record and legal authority are “complete and accurate,” warns of sanctions under Rule 11(c), and extends counsel’s confidentiality obligations to generative AI tools. It requires no disclosure of AI use. Other judges’ case-management orders were not swept one by one, so the assigned judge’s orders are worth checking at the start of each matter.

The D.C. Circuit’s Circuit Rules, as amended through December 4, 2025, contain no AI provision, and as of 2026-09-30 the court’s proposed-amendments page states that “No amendments are currently proposed.”

What DC local courts require for AI use in filings

No D.C. Court of Appeals or Superior Court rule requires AI disclosure or certification in filings as of 2026-09-30.

• Douglas v. Deutsche Bank (D.C. Ct. App. Sept. 3, 2026). In a published order in Douglas v. Deutsche Bank National Trust Co., No. 24-CV-1099, the court struck the appellee’s brief for fabricated citations under D.C. Appellate Rule 28 and referred the matter “to the Office of Disciplinary Counsel for any investigation it deems appropriate.” The court called it “the first reported case in which this court has received a brief containing fabricated citations,” and referred the scope of its sanctions authority to its Rules Committee.
• AI Task Force. By administrative order of March 6, 2026, the D.C. Courts made their Artificial Intelligence Task Force a standing committee. One of its listed purposes is to “Develop court rules with respect to disclosure, transparency, ethics, accuracy, authenticity and certification of AI use in court pleadings and proceedings.” The order itself imposes no duty on filers.
• Internal use policy. The D.C. Courts’ internal AI use policy was “promulgated on [05-16-2025]” and governs AI use by the courts’ judicial officers, employees, contractors, and volunteers. It is not attorney-facing. The courts’ AI page states that “Judges are not using AI to write opinions or make decisions.”

What AI-related rules are pending in DC?

Two court tracks could produce filer-facing rules, though neither had published a proposal as of 2026-09-30:

• The AI Task Force’s charge to develop rules on disclosure and certification of AI use in pleadings.
• The Rules Committee referral in Douglas on the scope of the Court of Appeals’ authority to sanction fabricated citations.

No enacted DC legislation governs attorney AI use. Mayor’s Order 2024-028 (February 8, 2024) set out AI values and established an advisory group for District government agencies; it does not apply to private law firms.

How do malpractice carriers in DC treat AI use?

No DC-affiliated bar carrier has published AI-specific underwriting positions in the materials reviewed for this entry. No filed AI exclusion on an admitted-market LPL form has been located. The claim patterns most likely to arise map to rules already in force: a fabricated citation in a filing (Rules 3.3 and 3.4, and now Douglas), and unauthorized access to client data through an AI vendor (Rule 1.6(f) and the Rule 5.3 service-provider comments).

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

No DC-dominant carrier has published AI-specific application items in the materials reviewed. Firms that want to be ready for questions may wish to have on hand a written AI use policy, vendor diligence records, a citation verification step, an AI billing practice consistent with Opinion 388’s time-based fee rule, a training log, and an incident response procedure for AI-related confidentiality events. Confirm application items directly with the carrier or broker.

What documentation should a DC firm keep on file?

Month one (foundational)

1. (Owner: managing partner + firm administrator) Written AI use policy covering permitted and prohibited tools, input handling, verification and review, AI billing, and approval authority. Opinion 388 describes firm-level measures of this kind as steps a firm could consider under Rules 5.1 and 5.3.
2. (Owner: litigation lead) Citation verification protocol for any filing that relies on AI-generated research: who verified each citation, against what source, and when. Rules 3.3 and 3.4 are the anchors, and Douglas shows the consequence of skipping this step in the D.C. Court of Appeals.
3. (Owner: matter lead attorney) Judge-specific AI order check at the opening of each D.D.C. matter, since chambers-level orders such as Judge Walton’s address generative AI even without a court-wide rule.

Months two and three (operational documentation)

1. (Owner: firm administrator + outside IT) AI vendor diligence record for each AI tool in use: privacy terms, data retention, whether client data is isolated, and the date of last review. This documents the reasonable care Rule 1.6(f) requires and the reasonable efforts Rule 5.3 Comments [3] and [4] describe.
2. (Owner: firm administrator) Competence training log recording attorney and staff training on AI tools, including date, content, and attendees. This supports the technology competence described in Rule 1.1 Comment [5].
3. (Owner: managing partner) Supervision protocol describing how partners confirm that supervised personnel use AI consistently with the Rules. Rules 5.1 and 5.3 are the anchors.

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) Per-matter AI use notes identifying which AI tool was used, the nature of its contribution, and the attorney’s verification steps. Opinion 388 identifies whether such interactions belong in the client file as a Rule 1.16(d) question.
2. (Owner: billing partner) AI billing practice. Under time-based fee agreements, time billed reflects the time actually spent, and direct AI tool costs pass through only where the client has agreed to pay out-of-pocket expenses (Opinion 388, Rule 1.5).
3. (Owner: managing partner) Incident response procedure for an AI tool that processes client confidential information in an unauthorized way, tied to the reasonable-care duty in amended Rule 1.6(f).

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