Most of the headlines about AI misuse in legal filings involve non-existent cases hallucinated by AI systems used in brief writing without cite checking. Attorneys relying on generative AI systems have been sanctioned after submitting briefs filled with invented citations (as discussed, for example, here). But a recent case out of Massachusetts shows that in using AI drafting can be a problem even when the cases are real.
A Boston lawyer is facing possible sanctions after submitting a filing that included 18 altered quotations from actual cases. This time, the AI system employed to improve the attorney's work appears to have also "polished” the quoted language and changed phrasing while keeping quotation marks intact.
This is thus another, albeit quieter but equally serious, risk in using AI for legal drafting. Tools marketed as grammar or style enhancers can also subtly (or not so subtly) rewrite language in quoted material, thereby misquoting the cited materials. The quotes look real and are from real cases, but they are no longer faithful to the original text.
The lawyer involved apologized and said he believed he was using the AI system like a glorified spell-check. But the court wasn’t satisfied, and the judge requested a revised explanation that counsel suggest an appropriate sanction, noting the time the court had spent chasing down errors.
This case marks a new entry in the growing list of AI-related cautionary tales for attorneys. Quote-checking is just as important as cite-checking, especially when drafting tools have been employed. A careful prompt might help avoid the issue, but in checking the final product, it's still necessary to verify every word in a quote as even the most familiar tools can betray trust.
This entire episode has been a debacle and embarrassment for counsel and appears to have been the lapse in oversight of using a tool . . . .
www.law.com/...
