Where a filing goes when you did not open the case
Lesson 5 of 5 · The attorney's week
1 · Learn the move · Red-team audit
The rules below do not ask whether the tool is good. They ask what you signed. Competence includes understanding the technology you use [M]. Candor to the tribunal makes a false statement of law yours, however it got into the brief [M]. The signature on a federal filing certifies that the legal contentions are warranted, and the court can sanction under the rule, under the statute on vexatious multiplication, and under its inherent power [M]. Supervising lawyers answer for the nonlawyers' use [M]. The ABA's formal opinion ties these together for generative AI and adds that a boilerplate consent is insufficient for a self-learning tool and that the bill carries time actually expended [P]. Judges have added standing orders that differ from courtroom to courtroom and change [S]; one circuit declined a court-wide certification rule and left the individual orders in place [P]. One state now has a statute on generative-AI use in legal work [S]; another has issued no formal opinion at the check date [S]. The sanctions orders that exist share one fact: the lawyer did not open the case [S]. The amounts, the dates and the case names live in the currency file and change; the rule does not. If you did not open the case and read the passage, it is not a citation yet.
Here are the rules I work under: [paste the rules above in your own words]. Here is one thing I did with AI this week, with what I pasted and what came back: [paste, de-identified now even if it was not then]. Act as the judge's clerk trying to find the line that gets me sanctioned. For every authority that came back, ask whether I opened it and read the passage. For every quotation, ask whether the words are on the page. For every item I pasted, ask whether it was information relating to a representation and whether the engagement named the tool. Ask whether I read the standing order on the filing date. Do not soften anything; only flag it. End with the one rule I was closest to breaking, as [VERIFY: rule number].
The facts this lesson rests on, dated
Each line names the source grade the lesson shows, [P] primary, [S] secondary, [M] unverified, and the day it comes due. Past that day, treat it as a question. Draft copy, not legal advice.
- Model Rules 1.1 cmt. 8 (technology competence), 1.4, 1.5, 1.6, 3.1, 3.3 (candor to the tribunal), 5.1 and 5.3 (supervision) and 8.4 are the text the sanctions orders cite; each state adopts its own version. [M] (researched Oct 6, 2026, not yet confirmed by a person, recheck by Dec 15, 2026). source
- Fed. R. Civ. P. 11, Fed. R. App. P. 38, 28 U.S.C. s. 1927 and the court's inherent power are the four sanction bases the AI-citation orders rest on. [M] (researched Oct 6, 2026, not yet confirmed by a person, recheck by Dec 15, 2026). source
- ABA Formal Opinion 512 (2024-07-29) applies Model Rules 1.1, 1.4, 1.5, 1.6, 3.3, 5.1 and 5.3 to generative AI: competence includes understanding the tool, a boilerplate engagement-letter consent is insufficient for a self-learning tool, and fees bill time actually expended. [P] (researched Oct 6, 2026, not yet confirmed by a person, recheck by Dec 15, 2026). source
- Judges' standing orders on generative AI differ by courtroom and change; a count of 143 such orders was reported in May 2026. Read the order in force on the filing date. [S] (researched Oct 6, 2026, not yet confirmed by a person, recheck by Dec 15, 2026). research pack C7 docket_lawyer.md (parity run 2026-10-05), citing the Federal Judicial Center
- The Fifth Circuit proposed, took comment on, and declined to adopt a court-wide generative-AI certification rule, leaving individual judges' standing orders in place. [P] (researched Oct 6, 2026, not yet confirmed by a person, recheck by Dec 15, 2026). source
- California SB 574, signed 2026-09-30 and effective 2027-01-01, adds disclosure requirements for generative-AI use in legal work; State Bar COPRAC guidance amendments were pending at the check date. [S] (researched Oct 6, 2026, not yet confirmed by a person, recheck by Dec 15, 2026). research pack C7 docket_lawyer.md (parity run 2026-10-05)
- No Tennessee formal ethics opinion on generative AI was found as of the check date. [S] (researched Oct 6, 2026, not yet confirmed by a person, recheck by Dec 15, 2026). research pack C7 docket_lawyer.md (parity run 2026-10-05)
- State opinions on generative AI include Florida 24-1, D.C. 388, Pennsylvania 2024-200, New York City 2024-5 and Texas 705; they differ on consent, disclosure and billing. [S] (researched Oct 6, 2026, not yet confirmed by a person, recheck by Dec 15, 2026). research pack C7 docket_lawyer.md (parity run 2026-10-05)
- A public database of court decisions involving AI-hallucinated material listed 2,149 cases as of 2026-10-05. [P] (researched Oct 6, 2026, not yet confirmed by a person, recheck by Dec 15, 2026). source
- Wadsworth v. Walmart (D. Wyo., 2025-02-24): $5,000 in sanctions and pro hac vice admission revoked over AI-fabricated citations in a filed motion. [S] (researched Oct 6, 2026, not yet confirmed by a person, recheck by Dec 15, 2026). research pack C7 docket_lawyer.md (parity run 2026-10-05)
- Johnson v. Dunn (N.D. Ala.): Butler Snow attorneys sanctioned and referred to the state bar over AI-fabricated citations. [S] (researched Oct 6, 2026, not yet confirmed by a person, recheck by Dec 15, 2026). research pack C7 docket_lawyer.md (parity run 2026-10-05)
- Whiting v. City of Athens (6th Cir., 2026-03): $15,000 per attorney over unverified AI-generated citations in an appellate brief. [S] (researched Oct 6, 2026, not yet confirmed by a person, recheck by Dec 15, 2026). research pack C7 docket_lawyer.md (parity run 2026-10-05)
- Couvrette v. Wisnovsky (D. Or.): $15,500 in sanctions plus $94,704.38 in fees, and the case dismissed with prejudice, over AI-fabricated authority. [S] (researched Oct 6, 2026, not yet confirmed by a person, recheck by Dec 15, 2026). research pack C7 docket_lawyer.md (parity run 2026-10-05)
- Reaves Law Firm v. Baker Donelson (W.D. Tenn., 2026-06-02): a Tennessee federal court's AI-citation sanctions ruling, the closest one to the firm's home courts. [S] (researched Oct 6, 2026, not yet confirmed by a person, recheck by Dec 15, 2026). research pack C7 docket_lawyer.md (parity run 2026-10-05)
- Proposed Fed. R. Evid. 707 would apply expert-testimony reliability standards to machine-generated evidence; proposed, not adopted, at the check date. [M] (researched Oct 6, 2026, not yet confirmed by a person, recheck by Dec 15, 2026). source
2 · Your turn. You write the prompt
The associate at Marchetti & Ostrow, honest in the retrospective: on Tuesday she pasted the client's internal emails into a consumer chatbot to draft the opposition, filed it Thursday with two cases she had not opened, and did not read the judge's standing order, which requires a certification. The partner wants the review done on this one filing before the firm's policy is signed.
Write the prompt for this scenario as given, or swap in a real one from your own work. The score works the same either way.
Remember: the AI sees only your prompt, not this page. If the situation isn't in your prompt, it doesn't exist.
Optional. These shape the output when you run your prompt below, not your score.