June 1, 2026 (in 3 days): New York: 22 NYCRR Part 161 takes effect, system-wide AI policy for all UCS courts

Virginia: AI Ethics Guidance for Law Firms

Formal opinion

Verified April 25, 2026

Citation
Virginia State Bar LEO 1901 (approved by Supreme Court of Virginia, Nov 24, 2025)
Opinion date
November 2025

Summary

Virginia has informal VSB guidance (Aug 2024) plus binding Legal Ethics Opinion 1901 on fees, approved by the Supreme Court of Virginia in November 2025. LEO 1901 expressly departs from ABA Formal Opinion 512 and North Carolina 2024 FEO 1: value-based fees are permissible even when AI dramatically reduces time, and attorneys are not required to reduce flat fees solely because AI was used. The verification duty is restated alongside this billing flexibility.

On this page
  1. Virginia State Bar LEO 1901 and informal guidance
  2. What federal courts in Virginia require for AI use in filings
  3. What Virginia state courts require for AI use in filings
  4. Has anyone been sanctioned for AI use in Virginia?
  5. What AI-related rules are pending in Virginia?
  6. How do malpractice carriers in Virginia treat AI use?
  7. What does my Virginia malpractice carrier ask about AI at renewal?
  8. What documentation should a Virginia firm keep on file?
  9. AI sanctions cases (7)
  10. Applicable rules (reference)
This summary is informational only. Verify the primary source before relying on this entry in any filing or client matter. Bar rules differ meaningfully by state; consult a licensed attorney in your state.

Bottom line for a 5-50 attorney Virginia firm: Virginia is the rare state where the binding bar opinion expressly permits value-based fees even when AI dramatically reduces time. LEO 1901 (approved by the Supreme Court of Virginia November 24, 2025) directly departs from ABA Formal Opinion 512 and North Carolina 2024 FEO 1 on this point, but it restates the verification duty in the same breath: no malpractice safe harbor for fabricated citations or unverified output. The biggest day-to-day exposure is the E.D. Va. patchwork, where individual judges (Young, Lauck, Novak, Vaala, Payne, Brinkema) issue case-specific AI orders that vary in scope.

Start with: a written AI use policy modeled on the VBA Model AI Policy, a citation verification log keyed to LEO 1901, and an E.D. Va. / W.D. Va. judge-specific AI order check at case intake.

Virginia State Bar LEO 1901 and informal guidance

Citation: Virginia State Bar Legal Ethics Opinion 1901, Reasonable Fees and the Use of Generative Artificial Intelligence. Approved by VSB Council June 12, 2025; approved by the Supreme Court of Virginia November 24, 2025. Final approved opinion (PDF). Proposed opinion (PDF).

Status: Formal Legal Ethics Opinion. Binding interpretive authority for Virginia attorneys under the Virginia Rules of Professional Conduct.

What it requires

  • Time-based fees may only reflect actual hours spent. A lawyer may not charge hourly time the lawyer did not spend.
  • Rule 1.5 permits non-hourly fee structures accounting for skill, judgment, experience, and results obtained, even when AI dramatically reduced time. A lawyer is not ethically required to reduce a flat or value-based fee solely because AI accelerated the work.
  • The “skill requisite to perform the legal service properly” may increase when using AI. Expertise in selecting, prompting, verifying, and integrating AI outputs is compensable.
  • “If counsel relies on artificial intelligence or other technology to draft a filing, the attorney is still responsible for ensuring the filing is accurate and does not contain fabricated caselaw or quotations.”
  • These principles extend to other technological tools producing comparable productivity gains.

What it recommends

  • When AI meaningfully increases efficiency, provide additional context in engagement communications. Explain why experience and technical skill contribute to the value of the fee, particularly under alternative fee arrangements.
  • Update engagement letters to reflect value-based fee rationale.

Notable divergences

LEO 1901 directly rejects the implication in ABA Formal Opinion 512 (July 2024) that efficiency-driven time savings require fee reductions. It departs from North Carolina 2024 Formal Ethics Opinion 1 on the same point. Virginia is explicitly more permissive on AI-enhanced billing than most state bars.

Notable gaps

LEO 1901 is focused on Rule 1.5. The opinion does not address which AI tools or vendor types are appropriate. Other unaddressed topics include client consent for confidential matters, data residency or cloud security standards, mandatory disclosure in court filings, conflicts between vendor terms and Rule 1.6, and AI use in intake or screening.

VSB informal guidance (August 2024)

Posted to the VSB Ethics and Conduct page (no opinion number; informal practice guidance, not binding precedent). Lawyers’ basic ethical obligations are unchanged by AI. Lawyers must vet how AI providers handle client data with the same diligence applied to any vendor. There is no per se requirement to inform a client about AI use unless the client agreement or matter sensitivity warrants it. Lawyers must review all AI output and verify that any citations are accurate and real. Rules 5.1 and 5.3 apply to AI-generated work. Attorneys must observe court rules regulating AI use.

VBA Model AI Policy (May 2024)

The Virginia Bar Association Task Force on Artificial Intelligence published a Version 1.0 Model AI Policy in May 2024. The policy covers tool selection and approval, training, data confidentiality, output validation, supervision, billing, and client communication. VBA is a voluntary membership organization. This template is persuasive, not binding. Model AI Policy (PDF). VBA AI resources page.

What federal courts in Virginia require for AI use in filings

No district-wide AI standing order exists in either E.D. Va. or W.D. Va. Individual E.D. Va. judges have issued case-specific AI orders that typically require disclosure of AI use and certification that all citations were independently verified. Requirements vary by judge. Judges with documented orders include Young, Lauck, Novak (in Dialect, LLC v. Amazon.com, Inc., No. 1:23-cv-581), Vaala (in Mwakalindile v. Truist Bank Corp., No. 1:25-cv-00112), Payne (in Powhatan Cnty. Sch. Bd. v. Skinger, No. 3:24-cv-00874, and Kromrey v. Capital One N.A., No. 3:24-cv-00575, with a carve-out for established legal research platforms such as Westlaw, Lexis, Fastcase, and Bloomberg), and Brinkema (in Hollinger v. Enlisted Ass’n of the Nat’l Guard of the United States, No. 1:25-cv-01349). Firms must check the assigned judge’s standing orders at the opening of each federal matter.

What Virginia state courts require for AI use in filings

Virginia has no state-court AI rule governing attorney filings. The Supreme Court of Virginia Rules for Generative AI Use govern judicial system employees, judicial officers, and court-issued device users. The rules prohibit autonomous AI decision-making and require human decision-maker involvement for AI-assisted judicial outputs. They do not directly regulate attorney conduct in filings.

Has anyone been sanctioned for AI use in Virginia?

In Iovino v. Michael Stapleton Associates, Ltd. (W.D. Va.), a party cited multiple cases and quotations that could not be located. The court described the situation as “ChatGPT run amok.” Virginia bar education materials frequently cite the case as the cautionary example that motivates the verification language in LEO 1901 and the VSB informal guidance.

HB 2094 (High-Risk AI Act, 2025 Regular Session) would have mandated AI risk disclosure in legal services, risk management plans, and algorithmic discrimination prevention. Governor Youngkin vetoed it on March 24, 2025. The veto was sustained April 2, 2025. No active attorney-conduct AI legislation is pending.

HB 1642 (Va. Code Ann. § 19.2-11.14), effective July 1, 2025, governs decisions on pre-trial detention or release, prosecution, adjudication, sentencing, probation, parole, correctional supervision, and rehabilitation. Such decisions must be made by a human decision-maker. Any AI-based tool recommendation must be subject to challenge or objection permitted by law. Defense attorneys may challenge AI-based tool recommendations used in criminal decisions. Firms doing criminal defense should understand AI deployments in local courts.

How do malpractice carriers in Virginia treat AI use?

The research note does not document a Virginia-affiliated carrier with published AI risk guidance. LEO 1901’s verification language and the VSB August 2024 informal guidance establish the substantive standard. A malpractice claim arising from AI-related error would be evaluated against that standard. LEO 1901 creates no safe harbor: value-based billing is permissible but attorneys retain full responsibility for output quality.

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

No Virginia carrier has published AI-specific renewal application items in the research record. The substantive standard the firm prepares against is LEO 1901 plus the VSB informal guidance: documented written policy, citation verification, vendor diligence, and court-specific disclosure where applicable. Confirm application items directly with the firm’s broker. Firms with criminal defense practices should additionally be ready to discuss their workflow for challenging AI-based tool recommendations under Va. Code § 19.2-11.14.

What documentation should a Virginia firm keep on file?

Month one (foundational)

Three foundational items.

  1. (Owner: managing partner + firm administrator) Written AI use policy structured on the VBA Model AI Policy: approved tools, data input restrictions, verification requirements, supervision protocols, and billing treatment. Anchors compliance with VSB August 2024 guidance and LEO 1901.
  2. (Owner: litigation lead) Citation verification log documenting that citations in any AI-assisted filing were independently verified before submission. Satisfies the verification duty restated in LEO 1901 and required by VSB informal guidance.
  3. (Owner: matter lead attorney) E.D. Va. / W.D. Va. judge-specific AI order check at case intake. Confirm whether the assigned judge has an AI disclosure order and log the result. If an order exists, file the required disclosure and certification.

Months two and three (operational documentation)

Two documentation items.

  1. (Owner: firm administrator + outside IT) Vendor due diligence record for each AI tool used on client matters. Cover data handling, confidentiality, retention, and self-learning status. Date and update annually or when terms change.
  2. (Owner: firm administrator) Attorney and staff training records showing the firm’s lawyers and staff have reviewed LEO 1901, the VSB August 2024 guidance, and the firm’s written AI policy. Supports Rule 5.1 and 5.3 supervision obligations.

Months four to six (per-matter discipline)

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

  1. (Owner: managing partner + billing partner) Engagement letter provisions for value-based or flat-fee matters. Explain that fees reflect the value of the work product, the complexity of the issues, and the results achieved. This is consistent with LEO 1901’s guidance on client communication under alternative fee arrangements.
  2. (Owner: matter lead attorney) Disclosure decision log recording, for each matter, whether AI was used and whether client disclosure was made. VSB informal guidance imposes no per se disclosure requirement, but logging the decision creates a defensible record.
  3. (Owner: managing partner) Periodic review schedule for new VSB guidance, new E.D. Va. and W.D. Va. judge orders, HB 1642 developments, and updated vendor terms.

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

AI hallucination sanctions cases in Virginia (7)

Editorially flagged cases for Virginia firms appear first with a "Why this matters" note; the remaining 4 entries collapse below.

Other Virginia cases (4)

Applicable rules (reference)

How each rule applies to AI use in legal practice. Rule numbers reflect this state's own numbering where it diverges from the ABA Model Rules.

Rule 1.1 : Competence
In the firm AI policy, list approved tools and document what each one can and cannot do, where its data goes, and where it hallucinates. The competence duty rests on the lawyer, not the vendor.
Rule 1.4 : Communication
Decide once, by matter type, when AI use rises to a level the client should be told about. Document the threshold in the firm AI policy and reflect it in the engagement letter.
Rule 1.5 : Fees
Bill actual time, not pre-AI hourly time. If AI subscription costs are passed through to a client, disclose the basis in writing before the invoice issues.
Virginia note: LEO 1901 (Nov 2025) expressly departs from ABA Op 512 and North Carolina 2024 FEO 1: value-based fees are permissible even when AI dramatically reduces time, and attorneys are not required to reduce flat fees solely because AI was used.
Rule 1.6 : Confidentiality
Vet every AI tool that touches client data: data retention, training-data use, security posture, breach disclosure, and deletion on termination. No client-identifying input into a public-tier tool.
Rule 3.3 : Candor Toward the Tribunal
Personally read every cited case and verify every quoted authority before signing a filing. This is the rule behind every AI-hallucination sanction issued to date.
Virginia note: LEO 1901 affirmatively restates the verification duty alongside the billing flexibility: no malpractice safe harbor, value-based billing does not relieve the attorney of accuracy obligations.
Rule 5.1 : Responsibilities of Partners and Supervisory Lawyers
Adopt a written firm AI policy, require AI training at intake, and verify compliance. Recent sanctions decisions (Mata, Johnson v. Dunn) have credited firms with pre-existing written policies as a mitigating factor.
Rule 5.3 : Nonlawyer Assistance
Treat AI tools the way the firm treats paralegals: written supervision protocol, attorney review before client or court delivery, and named accountability per matter.