In the matter of Machinsky

What the court wrote

Verbatim from the court's document (public domain). Ellipses mark omitted text; nothing is paraphrased.

In addition to this false quotation, a review of other legal citations in defendant's filings uncovered several additional inaccuracies and questionable string citations that may be attributable to use of generative artificial intelligence (“AI”) tools or large language models (“LLMs”). On page 3 of defendant's Motion to Recuse, he writes: “The appearance of bias or improper ex parte communications is sufficient to require recusal, even absent a showing of actual bias. Commonwealth v. Eddington, 71 Mass. App. Ct. 138, 144-145 (2008); Commonwealth v. Gogan, 389 Mass. 255, 259 (1983); Commonwealth v. LeBlanc, 475 Mass. 820, 822-823 (2016).” While Eddington and Gogan address the issue of recusal, both affirm that recusal was not required under the applicable standard and circumstances. However, LeBlanc is entirely irrelevant. LeBlanc concerns whether leaving the scene of property damage requires proof that the accident occurred on a public way. It does not mention recusal at all. ... The presence of fabricated quotations and inaccurate citations strongly suggests that defendant used unverified or unreliable sources, such as AI tools or LLMs (like ChatGPT, Claude, Gemini, or Copilot) when drafting his filings. While this does not bear directly on the outcome of this S-case, the court documents this issue on the record to serve as a warning to defendant and other litigants (and an alert to judicial officers in related proceedings) about the serious risks of AI-generated “hallucinations” and false citations when AI tools or LLMs are used without adequate oversight.

Authorities cited in the passage: Commonwealth v. Eddington, 71 Mass. App. Ct. 138 (2008) · Commonwealth v. Gogan, 389 Mass. 255 (1983) · Commonwealth v. LeBlanc, 475 Mass. 820 (2016) · Mass. R. Civ. P. 70 · Peterson v. Hopson, 306 Mass. 597 (1940) · King v. Driscoll, 424 Mass. 1 (1996)