Updated
November 16, 2026

The Reasonable Machine

Tracking artificial intelligence in the law

Where AI in the law stands, and what to expect in 2027

Courts have now issued more than 2,200 decisions dealing with citations or quotations that an AI tool made up. A federal judge has ruled that a client's chats with a public AI tool were not privileged. New AI laws and rules in four states start on January 1, 2027.

The law measures conduct against a reasonable person. This page follows what that standard requires now that software does part of the work. Every item links to its source, and forecasts are marked as forecasts.

Latest, November 13, 2026

Delaware's Court of Chancery sanctions a law firm partner over an AI-assisted brief

The court ordered a partner at one of Delaware's large firms to pay $2,500 for an answering brief that contained AI-generated errors he had not checked. The firm had a written AI policy and required ethics training. The court said generative AI is allowed in legal practice and that the care its use requires was missing. Bloomberg Law has the report.

Where things stand

Made-up citations

The count has passed 2,200 decisions and the penalties are larger

Researcher Damien Charlotin keeps a public database of court decisions involving AI-invented material. It listed 2,262 decisions worldwide on November 2, 2026, with 1,533 from United States courts. Many involve people without lawyers.

The first widely reported case, in 2023, ended in a $5,000 sanction. In 2026 two Oregon lawyers were ordered to pay about $110,000 in sanctions and fees and their client's case was dismissed. The Nebraska Supreme Court suspended a lawyer over a divorce brief, and a federal court of appeals has ordered $15,000 per attorney. The Wisconsin Law Journal and a table of 2026 cases collect the details.

Privilege

A federal judge held that chats with a consumer AI tool were not privileged

In United States v. Heppner, decided in February 2026, Judge Jed Rakoff of the Southern District of New York ruled that about 31 documents a criminal defendant created with a consumer AI chatbot were protected by neither attorney-client privilege nor the work product doctrine. The chatbot is not a lawyer. The defendant also used it on his own, without direction from his lawyers. Baker Botts and Bloomberg Law explain the ruling.

Courts have not all gone the same way. A federal court in Michigan treated a self-represented person's ChatGPT conversations as work product, according to MIT Technology Review.

Discovery orders

Protective orders have started to say which AI tools may touch case material

Protective orders entered in Southern District of New York criminal cases in August, September and October 2026 include a section restricting the use of AI tools on discovery material. One order requires defense counsel to be able to certify, after checking the provider's terms and account settings, that the tool keeps the material confidential. Examples are United States v. Stamp (August 11, 2026) and United States v. Mendivil (October 5, 2026).

People without lawyers

Self-represented filings in federal court are up sharply

A study of 4.5 million federal civil cases found that the share brought by self-represented people rose from 11% in 2022 to 16.8% in 2025, MIT Technology Review reported. Complaints showing signs of AI drafting went from almost none before 2022 to about 18% of self-represented complaints in early 2026. The outcomes have not moved. Self-represented litigants still win fewer than 1% of their cases, according to Law360.

Evidence

The proposed federal rule on AI evidence is on hold

Proposed Federal Rule of Evidence 707 would hold machine-generated evidence offered without an expert witness to the reliability standard that applies to expert testimony. Public comment closed on February 16, 2026. At its May 7, 2026 meeting the Advisory Committee on Evidence Rules decided not to move the rule forward and kept it under study. A separate proposal on deepfakes, Rule 901(c), has not been published for comment. See Baker Botts and the Center for Democracy and Technology.

Ethics rules

Most states have guidance, and California is moving toward binding rule text

Most states have dealt with lawyers' AI use through ethics opinions and guidance. California's ethics committee approved proposed amendments to six Rules of Professional Conduct in March 2026 and a second round in June. They would bind lawyers only after the State Bar's Board of Trustees approves them and the California Supreme Court adopts them.

The New York State Bar Association adopted AI ethics guidelines on October 31, 2026. Oregon issued an opinion on AI agents that handle client intake.

Judges

Most surveyed federal judges have used an AI tool

In a Northwestern University survey, more than 60% of 112 responding federal judges said they had used at least one AI tool in their judicial work, the New York State Bar Association noted. A federal judiciary task force led by Judge Sidney Thomas of the Ninth Circuit aims to release final guidance on courts' use of AI by the end of 2026, the ABA reported.

The business

Lawyers use the tools weekly, and firms have been slower to set rules

Seventy percent of law firm attorneys use AI at least once a week, according to Law360 Pulse's 2026 survey. The 8am 2026 Legal Industry Report found that 69% of legal professionals use general-purpose AI tools for work, up from 31% a year earlier. It also found that 54% of firms offer no AI training and 43% have no AI policy.

About 90% of legal spending still runs through hourly billing, Thomson Reuters reports. Investors keep funding the vendors. Harvey announced a $550 million round at a $15.5 billion valuation on September 9, 2026.

Legislation

The states are writing the AI rules and Congress has not overridden them

No federal statute preempts state AI laws. A December 2025 executive order asserted federal authority and directed challenges to state laws, and Congress has so far declined to pass a preemption statute, Ropes & Gray explains. A preemption tracker found no such statute as of November 4, 2026.

Dates already set

The 2027 calendar

End of 2026, expected
Final guidance from the federal judiciary's AI task force on how courts use AI.
January 1, 2027
Colorado's rewritten automated decision-making law, SB 26-189, and its chatbot safety law take effect.
New York's RAISE Act and Illinois SB 315, both aimed at developers of the most advanced AI models, take effect.
California's privacy rules on automated decision-making technology begin to apply.
December 2, 2027
European Union AI Act obligations for stand-alone high-risk systems begin, moved from August 2026.
August 2, 2028
European Union obligations for high-risk AI built into regulated products begin.
No date yet
Federal Rule of Evidence 707, California's AI amendments to its ethics rules, and Utah's rule on AI legal tools.

State dates are from the George Mason University 50-state tracker and Jenner & Block. European dates are from Gibson Dunn and the Future of Privacy Forum.

What to expect in 2027

  • Scheduleda date is set in law
  • Likelyour forecast from the current trend
  • Opencould go either way
State law

ScheduledState AI laws start on January 1 with no federal override

The Colorado, New York and Illinois statutes and California's automated-decision rules begin on the first day of the year. Lawyers at Foley & Lardner, writing after the November election, expect the push for a congressional override to stall. Businesses should plan around the state rules.

Sanctions

LikelyMore sanctions, at larger firms, in larger amounts

The United States count in the public database went from 808 in March 2026 to 1,533 on November 2. Firm policies have not stopped it. The Delaware lawyer sanctioned on November 13 worked at a firm with an AI policy and required training. Courts have been raising the cost each year and we expect that to continue.

Privilege

OpenAn appeals court may take up AI chats and privilege

Heppner is one trial judge's ruling, and at least one other trial court has protected a self-represented person's AI work. We have not seen an appeals court decide the question. Until one does, the cautious assumption is that anything typed into a consumer AI tool can be obtained by the other side.

Ethics rules

LikelyAI duties move from guidance into rules lawyers can be disciplined under

California's six proposed amendments would write AI duties into the text of its Rules of Professional Conduct. If the Board of Trustees and the California Supreme Court act in 2027, other states will have a model to copy.

Courts

LikelyMore court orders that name the AI tools lawyers may use

The federal judiciary's guidance is due, and trial judges are already writing AI terms into protective orders. Expect more standing orders on disclosure and more orders limiting which tools may process discovery. The University of North Carolina law library keeps a guide to existing court guidance.

Evidence

LikelyAI evidence and deepfake disputes get decided under the existing rules

With Rule 707 and Rule 901(c) still in committee, judges will keep applying the authentication and expert testimony rules they already have. Lawyers offering or attacking AI-generated evidence should build the record under those rules.

Fees

LikelyClients press harder on price

A quarter of firms in BigHand's 2026 pricing research already discount when a client permits AI use. In a survey of large-firm innovation leaders published by Harvey, a vendor, 73% said clients increasingly pick firms based on their use of AI. Thomson Reuters' 2027 rates report says firms have not yet cut junior staffing in any substantial way and that further automation could put pressure on hourly pricing.

Common questions

Are my chats with ChatGPT or Claude protected by attorney-client privilege?

Assume they are not. In United States v. Heppner, a federal judge in New York ruled in February 2026 that documents a defendant made with a consumer AI chatbot were not privileged and were not work product. The judge noted that the chatbot is not a lawyer and that the defendant used it without his lawyer telling him to. A federal court in Michigan reached a different result for a person representing herself. We have not seen an appeals court decide the question. If you have a lawyer, ask before you put facts about your case into any AI tool.

Can my lawyer use AI on my case?

Yes, in general. Courts that have sanctioned lawyers have said the use of AI was not itself the problem. Bar guidance in California, New York and other states says the ordinary duties still apply. The lawyer has to protect your confidential information, check everything the tool produces, and stay responsible for the work.

Do I have to tell the court if I used AI to write a filing?

It depends on the court. Some federal districts and individual judges require a disclosure or a certificate that a person checked the citations. Many courts have no AI rule. In every court, whoever signs a filing is responsible for what it says. Read the court's local rules and the assigned judge's standing orders before you file.

Can I use AI to represent myself in court?

You can, and many people do. About 18% of complaints filed by self-represented people in federal court in early 2026 showed signs of AI drafting. The results have not improved. Self-represented litigants still win fewer than 1% of their federal cases. Courts hold people without lawyers responsible for made-up citations too, and the public database includes many warnings and penalties against them.

Will AI replace lawyers?

Not on the evidence so far. Thomson Reuters reports that firms have not substantially cut junior staffing. In a Citi and Hildebrandt survey, 86% of large firms planned to grow their associate ranks through 2027, and 35% planned larger first-year classes. Firms are also hiring experienced lawyers who know the tools ahead of new graduates.

Close to home

Utah

May 22, 2025

The Utah Court of Appeals sanctioned lawyers for AI-invented citations

In Garner v. Kadince, Inc., the court addressed fake citations in a filing that appeared to have been written with AI. It was the first time a Utah court had dealt with AI in the preparation of court papers. The sanctions included a $1,000 payment to the legal aid nonprofit "and Justice for all." KSL and the Deseret News covered the decision.

2026 and 2027

OpenThe Utah Supreme Court's committee is drafting a rule for AI legal tools

The Supreme Court's Ad Hoc Committee on Regulatory Reform has an AI and legal technology workgroup. Its question is when software that helps people with legal problems counts as the practice of law. The committee reported in March 2026 that the Court was receptive to a possible carve-out in the unauthorized practice rules. In May 2026 the workgroup agreed to proceed by rule and began drafting.

The same minutes note that Utah rule changes are usually published on the first of May and the first of November. A rule in 2027 is possible and no date is set. The workgroup is not reviewing how lawyers or courts use AI. Bar committees and the Administrative Office of the Courts are studying those questions.

Sources

Figures are as of the date each source gives. This page was last checked against them on November 16, 2026.

  1. AI Hallucination Cases Database, Damien Charlotin
  2. Delaware Chancery Court Sanctions Attorney for AI Hallucinations, Bloomberg Law, November 13, 2026
  3. AI-generated errors persist in court filings, Wisconsin Law Journal, November 9, 2026
  4. AI Hallucination Sanctions in 2026, Legal Desire, November 3, 2026
  5. What United States v. Heppner Means, Baker Botts
  6. Heppner Shows Attorney-Client Privilege's Fragility in AI Era, Bloomberg Law, March 10, 2026
  7. How courts are coping with AI-generated lawsuits, MIT Technology Review, June 4, 2026
  8. AI Drives Spike In Pro Se Filings But Not Courtroom Success, Law360
  9. Federal Evidence Rulemaking on AI Hits Pause, Baker Botts, June 2026
  10. The Advisory Committee Reviews Its AI Evidence Rule, Center for Democracy and Technology
  11. Proposed Amendments to the Rules of Professional Conduct Related to Artificial Intelligence, State Bar of California
  12. New York Bar Issues AI Guidelines, FindLaw
  13. New Oregon State Bar Opinion on Using AI Agents for Client Intake, NWSidebar, March 23, 2026
  14. AI and the Courts, New York State Bar Association, August 12, 2026
  15. Federal judiciary plans to release AI recommendations, American Bar Association, October 15, 2026
  16. The 2026 AI Survey, Law360 Pulse, March 31, 2026
  17. Generative AI data from the 2026 Legal Industry Report, 8am
  18. When AI efficiency collides with billable time, Thomson Reuters
  19. Law Firm Rates Report 2027, Thomson Reuters Institute
  20. Law Firms' Junior Roles At Risk From AI, Report Says, Law360 Pulse
  21. State AI Policy: A 50-State Tracker, George Mason University
  22. State Tech and AI Regulation in 2027, Jenner & Block
  23. What Democratic Control of Congress Means for AI Regulation, Foley & Lardner
  24. The Federal Push to Override State AI Regulation, Ropes & Gray, March 2026
  25. EU AI Act Omnibus Agreement, Gibson Dunn
  26. The AI Act Implementation Timeline, Future of Privacy Forum
  27. Supreme Court's Ad Hoc Committee on Regulatory Reform, Utah State Courts
  28. Garner v. Kadince, Inc., Utah Court of Appeals, May 22, 2025; United States v. Heppner, Southern District of New York, February 17, 2026; United States v. Stamp and United States v. Mendivil, Southern District of New York, 2026