In an AI legal clusterf*ck last month, a federal judge in Mississippi canceled a trial, removed all four lawyers from a civil case over legal fees, suspended two of them from appearing in that court for two years, and imposed fines after discovering that both sides had relied on AI tools that produced fictitious case citations in their filings.
A few details from the latest news coverage are worth noting:

1. One attorney reportedly told the court she was “unaware that A.I. could produce hallucinated cases” and “did not even know what a hallucinated case was.”
2. She wasn’t hacking together prompts on a hobby chatbot; she was using commercial software marketed specifically to help attorneys draft legal documents.
3. Neither side caught the other side’s bogus citations or faulty arguments, even after the judge raised concerns and ordered them to explain their AI use.
4. The underlying dispute was about legal fees, which makes the whole episode even more ironic: lawyers fighting over what their work is worth, while demonstrating how little value unverified AI output can have in a courtroom.

Stories like this are no longer rare. Reports of AI‑generated “hallucinated” case law, fake quotations, and fabricated authorities in court filings are surfacing almost daily.
We should be clear in how we talk about this: when licensed professionals sign their names to documents containing made‑up citations they never bothered to verify, that isn’t a tech glitch or a harmless experiment. It is, in plain terms, legal malpractice. And it’s a warning to every profession, including accountants and other advisers, that blindly delegating judgment to AI is not just sloppy; it’s sanctionable.
This isn’t just a lawyer problem. In tax and accounting, we are only one sloppy prompt away from the same kind of public embarrassment and possible financial damages. When CPAs use AI to draft memos, prepare opinion letters, or help assemble filings, the duty is identical: we must verify every authority, every citation, and every material fact before we sign. The most common AI‑related problem I face lately is when clients do their own online research and compile “tax evidence” that does not fit their case. I often have to say, “I know what you read online, but in this case…”
Courts are already saying that “I didn’t know AI could hallucinate” is not a valid excuse. It is incompetence. If we let unverified AI output slip into returns, protests, or financial statements, we’re not “experimenting with new technology”; we’re taking malpractice‑level risks with clients’ financial lives and our own professional reputation.