The judicial branch published several amendments to the Connecticut Practice Book that take effect immediately, including a new Section 4-9 titled "Generative Artificial Intelligence ('Generative AI') Compliance." The rule says anyone who files papers in court is representing "that they have made good faith, diligent efforts" to comply with every professional obligation and state law regarding the use of AI tools.
"Due to the risk that generative AI can create inaccurate factual and legal information, including, without limitation, faulty citations to legal authority, fabricated quotations from such authority and inaccurate or fabricated evidence, any person who uses generative AI in the creation or editing of any document filed with the court shall independently verify all citations, legal authorities or evidence produced by generative AI," the rule says. "The failure to do so may result in court-imposed sanctions, including, without limitation, the entry of a nonsuit or default judgment."
The rule defines generative AI as encompassing "ChatGPT, Co-Counsel, Google Bard, Neeva, Harvey, Ironclad, Bing, DeepSeek, Grok, and similar technology," and it specifies that compliance is solely the responsibility of the filer.
"The court or the clerk of the court is not required to review any filed document for compliance with this rule," the rule says.
The commentary for the new rule says the technology "may offer certain efficiencies in drafting legal documents" when used properly, but also carries "inherent risks."
Courts in Connecticut and all over the U.S. regularly identify apparent AI hallucinations in filings, sometimes leading to sanctions for those responsible for verifying the contents.
Greenwich solo practitioner Hilary B. Miller is facing possible sanctions for two documents he submitted in a taco restaurant trademark battle that were sullied by artificial intelligence errors. He told a federal judge in a written declaration Thursday that he was "mortified" by the mistakes and that he understands now that the duty of verification "requires that every quoted passage be exact, every [pinpoint citation] correct, every parenthetical a fair statement of what the authority holds or supports, and every paraphrase of law faithful to its source."
In February, a different Connecticut federal judge ordered employment attorney David Stich to undergo three hours of continuing legal education courses on responsible AI use and share what he learned with other lawyers. Stich filed a brief that included AI-generated quotes, the order said; the judge's decision came a few months after he was ordered to pay $500 for filing fake case citations with the court.
In September, federal judges in Connecticut started adding warnings to new case dockets that let litigants and their counsel know about a zero tolerance policy for AI hallucinations.
The Practice Book amendments also include a revision to Section 4-2(b), which governs the signing of documents like pleadings, motions and requests. A signature will now act as certification that the signer has complied with the AI rule in addition to requirements for shielding personal information.
The Rules Committee of Superior Court voted June 11 to accept a raft of proposed amendments. Besides the AI rules and a new Section 31a-18A, dealing with the transfer of children on probation to adult probation, they all take effect Jan. 1.
The new Section 3-4A allows an attorney or party to file a notice of preferred pronouns/salutations to designate a form of address.
"When a notice of preferred pronouns/salutations is brought to the attention of the judicial authority, or the judicial authority is otherwise alerted to an individual's preferred pronouns or salutation during any court proceeding, the judicial authority shall endeavor to use the individual's name, the designated salutation or personal pronouns, or other respectful means that are consistent with the individual's designated salutation and personal pronouns when addressing, referring to or identifying the individual party or attorney," the rule says.
According to commentary on the rule, judges are "required" to be courteous, and that includes "treating individuals in a manner that is objectively respectful of gender identity."
"This rule is intended to be directory rather than mandatory, and the rule is not intended to prohibit the inadvertent or mistaken misgendering of an individual by the judicial authority," the commentary says.
--Additional reporting by Emily Sawicki. Editing by Karin Roberts.
For a reprint of this article, please contact reprints@law360.com.
