The rules are changing under you: AI requirements lawyers need to track

If you practice in more than one state, or in federal court, you are already dealing with a patchwork. It is moving, so here is a snapshot. Verify anything you rely on against the source, because this changes fast.

In force

Florida Rule 2.515 (effective June 15, 2026). The person who signs a filing certifies that the authorities cited exist. This turns what used to be an inference from Rule 11 into an explicit statement.

California Rule of Court 10.430 (effective September 1, 2025). Courts that allow judicial officers and staff to use generative AI must have a use policy, and the policy must bar entering confidential or nonpublic information into public AI tools and require reasonable steps to verify accuracy. It applies to courts and court staff, not to lawyers' filings. But it tells you what judges are being told.

Ethics opinions. ABA Formal Opinion 512 (July 2024) and state opinions, including Florida 24-1, Texas 705, Virginia LEO 1901 and California's practical guidance, apply existing duties to AI: competence, confidentiality, supervision, candor and fees. They are not statutes, but bar counsel reads them.

Enacted, not yet effective

California SB 574. As summarized by law firm alerts, it was signed September 30, 2026 and takes effect January 1, 2027. Reported provisions include: lawyers may not delegate the practice of law to generative AI; confidential or personally identifying information may not be entered into a generative AI system unless access is limited to the lawyer and authorized people bound by confidentiality; lawyers must take reasonable steps to verify outputs; and filings may not contain citations the lawyer has not personally verified. The disclosure requirement to courts appears to be broad, but its scope was described as unclear. Read the enacted text before you build a policy around a summary.

Proposed

California Rules of Professional Conduct. The State Bar's committee approved proposed changes in March 2026 and the public comment period closed in May 2026. They would add AI-related comments to Rules 1.1, 1.4, 1.6, 3.3, 5.1 and 5.3. They were not final as of the sources we reviewed, and Supreme Court approval is required.

Federal Rule of Civil Procedure 11. There is a pending suggestion to amend Rule 11 for AI. It is a suggestion, not an adopted change.

Stalled

Proposed Federal Rule of Evidence 707 (machine-generated evidence). The advisory committee published it for comment in August 2025, held hearings in January 2026, and in May 2026 voted not to advance it or three redrafts for republication. A separate deepfake proposal on Rule 901 was postponed. The earliest republication discussed was August 2027. We could not confirm the outcome of the committee's October 2026 meeting. Do not tell clients there is a federal AI evidence rule. There is not one.

Courts are choosing not to write new rules

In Fletcher v. Experian, the Fifth Circuit declined to adopt a local AI-certification rule, saying existing obligations already cover it. Expect more of this: judges saying the rules you already have are enough.

Fees

No court decision on AI-assisted billing turned up in our research. The guidance is consistent, though. ABA 512 says hourly bills must reflect actual time, efficiencies should show up in flat fees, and general AI subscriptions are overhead. Virginia LEO 1901 says reduced time does not automatically mean a proportionally reduced fee and that value-based fees must be explained to the client. California's guidance says markups on matter-specific AI costs need informed written consent.

Insurance

Malpractice carriers are watching. One carrier's commentary, as reported in October 2026, was that AI shows up as part of the fact pattern in claims rather than as a wave of direct AI claims, and that policies generally do not exclude negligence arising from generative AI use. Carriers expect to ask more detailed underwriting questions. Check your own policy and your own application.

What to do with all of it

You cannot build a compliance program around every proposal. Build it around the duties that are not going anywhere:

1. Verify every citation before it goes out the door.
2. Protect client information, and know where it goes.
3. Supervise what lawyers and tools under you produce.
4. Be able to show your work to a court, a client or a carrier.
5. Bill honestly for AI-assisted work.

JILL is designed to support those duties. Citations are validated before they reach a draft, paragraphs carry confidence scores, matters are walled and access-controlled, firm data is not used to train third-party models, and every action is recorded in the DDEAS audit log. When a rule changes, the underlying need stays the same: be able to show what you did and why you can stand behind it.

Built for ABA Formal Opinion 512 →

50-state AI guidance survey →