$1,500 Per Fake Citation: What the First District Just Told Every Chicago Law Firm

/blog/illinois-appellate-court-ai-sanctions-chicago-law-firms

On July 28, the Illinois Appellate Court for the First District fined a Chicago attorney $15,000 and referred him to the ARDC. The math is what should get your attention: the court assessed $1,500 for each of ten false citations and statutory quotations, and said plainly that it was pricing above what other courts had imposed "because the sanctions being handed out so far weren't working.

The attorney wasn't a solo practitioner cutting corners. He was using what he described as a premier corporate ChatGPT subscription to synthesize complex matters, and he told the court he cross-referenced citations against LexisNexis to avoid hallucinations. He had a process. The process failed. Worse, his response brief defending the original filing contained its own fabricated statutory subsection.That last detail is the one I keep coming back to. This wasn't one bad afternoon. It was a workflow that produced errors, and then produced more errors while explaining itself.## The Real Failure Wasn't the ToolEvery managing partner I talk to in Chicago has a version of the same reaction to these cases: our people know better. Maybe. But look at what actually broke here.

The attorney used a general-purpose chatbot for legal research. ChatGPT is not a legal research tool. It has no connection to Westlaw, Lexis, or any authoritative case database. When you ask it for supporting authority, it generates text that resembles a citation, because generating plausible text is the entire function. It is doing exactly what it was built to do. The failure is in the assignment, not the tool.Then the verification step collapsed under time pressure. He said he cross-referenced. Ten citations got through anyway. Verification that depends on an individual attorney's discipline at 11 p.m. before a filing deadline is not a control ” it's a hope.And there was no second set of eyes. In a firm with a real AI policy, work product touching AI-generated research gets a defined check before it leaves the building. In most firms, it doesn't.

A 2026 industry survey found 43% of legal professionals say their firm has no formal AI policy and no plans to write one, and only 9% report a written policy that is actually enforced. The exposure is not theoretical.

Illinois Gives You Room” and That's the TrapHere's what makes this a specifically Illinois problem. The Illinois Supreme Court's AI policy, effective January 2025, is among the most permissive in the country. It does not require attorneys to disclose AI use. It does not restrict which tools you use. It says the Rules of Professional Conduct apply in full, that you must review AI output for accuracy before submission, and that confidential information must be protected.That's a policy built on trust. It gives Illinois firms more operating freedom than attorneys in states with mandatory disclosure certifications. It also means nothing external will catch your mistake before a judge does. There's no checkbox on the filing, no certification prompt, no procedural speed bump. The First District is now supplying the accountability the front end doesn't and it has explicitly signaled that fines will keep climbing until they deter.Permissive rules plus no internal controls is the worst combination available.

If your firm has been treating the Supreme Court's policy as permission to skip building a governance framework, that reading just got expensive. Three Things to Fix This Week**Separate research from drafting.** Draw a hard line: general-purpose AI never touches legal research. Full stop. If you want AI in research, it has to be a tool grounded in an actual legal database with linked, verifiable citations” Clio's Vincent, vLex, Lexis Protege, CoCounsel.

For transactional work, tools like Spellbook operate on the document in front of you rather than inventing authority. That's a categorically different risk profile.**Make verification structural, not personal.** Every citation in a filing gets pulled and read by a human, and someone signs off that it happened. Put it in the file.

When it lives on a checklist instead of in an attorney's memory, it survives deadline pressure.**Write the policy down.** One page. Which tools are approved, what they may be used for, what client data may never enter them, who reviews AI-assisted work. The ARDC's Illinois Attorney's Guide to Implementing AI is a reasonable starting frame. A policy nobody has written is a policy nobody follows.None of this slows a competent firm down.

The firms getting real leverage from AI in Chicago right now ”and they exist” are the ones that built these guardrails first, then moved fast inside them. The guardrails are what make speed safe.

If you're not certain what your attorneys are pasting into which tools, you have a governance gap, and it costs $1,500 a citation to find out the hard way. As an authorized Clio and Spellbook reseller, jamesmadison.ai can demo both tools in a single call and show you what grounded, citation-verifiable legal AI actually looks like next to a general-purpose chatbot. [Book a free 30-minute discovery call at jmaifirm.com.]---*jamesmadison.ai advises Chicago-area law firms on AI adoption, implementation, and governance. Our AI Adoption Workshop includes 3.75 hours of Illinois CLE credit.*