Discover Docket
← The DocketAI SANCTIONS

The second signature problem: why supervising lawyers keep getting sanctioned

Your name on the brief means you checked it. Courts are treating that as a fact, not a formality.

DDDiscover DocketOctober 27, 2026 · 7 min read

Most firms run on the same unwritten rule. The associate drafts. The partner reviews. The partner signs.

That worked because the partner could see the associate's work. The associate cited a case, the partner could pull the case. The failure mode was visible.

AI breaks that. The draft shows up looking finished, with citations in the right format and confident language. Nothing in it tells the reviewer where the weak spots are. Supervision becomes a trust exercise, and the sanctions orders show how that goes.

What the courts have actually done

Wadsworth v. Walmart (D. Wyo., Feb. 2025). A motion in limine contained eight fabricated citations. The lawyer who drafted it using the firm's in-house AI platform was fined $3,000 and lost pro hac vice admission. The two lawyers who signed the motion were fined $1,000 each under Rule 11 for signing without a reasonable inquiry. The firm itself was not sanctioned, in part because it had taken steps on verification.

Lacey v. State Farm (C.D. Cal., special master, May 2025). $31,100 imposed jointly on two firms. About 9 of 27 citations in the brief were wrong. When a brief passes between firms, nobody owns the check, so everybody owns the sanction.

Johnson v. Dunn (N.D. Ala., July 2025). This is the one every supervising partner should read. The firm had a written AI policy, had warned lawyers in 2023, and audited more than 2,400 citations once the problem surfaced. The firm was not sanctioned. The lawyers were. A public reprimand, disqualification from the case, and a referral to the state bar. The partner whose name stayed on the signature block without a check was in the group that got sanctioned.

Gordon Rees (Bankr. M.D. Ala., Nov. 2025). A large firm paid the opposing side's fees and circulated new policies internally after fabricated citations appeared in a filing.

What these have in common

Rule 11 and the supervision rules (Model Rules 5.1 and 5.3) do not ask whether you personally typed the fake citation. They ask whether you made a reasonable inquiry before putting your name on it, and whether you had measures in place to make sure the lawyers and tools under you were doing their job.

ABA Formal Opinion 512 says it directly: managerial lawyers need to establish clear policies on AI use and make sure the people and tools they supervise comply. Good intentions do not satisfy that. You need to be able to show what was done.

And a policy alone, as Johnson shows, protects the firm but not the individual.

Supervising what you cannot see

The practical problem is that you cannot supervise a process you cannot inspect.

If your associate used an AI tool in a browser tab, you have no record of what they asked, what came back, what they changed, or what they never checked. The only evidence is their word.

This is where the architecture matters more than the policy.

JILL works inside the matter. The request, the sources retrieved, the draft produced and the confidence score on each paragraph are all tied to the matter record, not to someone's personal browser history.

The audit log is the supervisor's window. Each action is recorded in an append-only, signed, chained log under DDEAS. A supervising lawyer can see what was generated and what it was based on. The review stops being "I trust that you checked" and becomes "I can see what was checked."

Citations are validated before they reach the draft. The person signing is not the first line of defense against a fake case. The system is.

Ethical walls and role-based access follow the matter. The right people see the right matter. That matters when you are trying to show who had access to what.

What this does and does not do

To be clear: JILL does not sign for you, and it does not make the supervision duty go away. You still have to read the work and form a judgment. What it does is give the supervisor something to supervise: a validated draft, a signal on where it is weak, and a record of how it got made.

A name on a signature block should mean something. In 2026, courts are treating it as meaning the lawyer actually checked. Build the practice so it does.

See how the audit trail works →

Learn about JILL →

Continue reading

AI SANCTIONS

ABA Formal Opinion 512: The lawyer's new duties in the AI era

In July 2024 the ABA issued Formal Opinion 512 — the most consequential ethics guidance of the AI era. It identifies six distinct duties every lawyer using generative AI now owes. Here's what each one requires.

Discover Docket · June 9, 2026 · 10 min read

Stop running your firm on fifteen tools.

Discover Docket replaces case management, research, AI, depositions, billing, and communications in one platform. California and Federal first, 52 jurisdictions on day one.