Tov Realty, LLC v. Suarez, 355 Conn. 902 (2026).
Background: Attorney Ian G. Gottlieb and his firm, GLG Law, LLC, filed applications for certification to appeal and subsequent briefs in two consolidated cases (Tov Realty v. Suarez and Kosel Equity v. MacGregor). Both filings contained hallucinated case citations generated by ChatGPT—roughly seven erroneous or unverified citations in total.
Where the lawyer went wrong:
- Verified once, then let AI touch the work unchecked. Gottlieb did his initial legal research properly, using LexisNexis and Shepard’s to confirm the citations in his rough drafts were accurate. His mistake was pasting those already-verified drafts into ChatGPT afterward, purely to improve organization and prose quality.
- Assumed the AI wouldn’t alter substance. He didn’t anticipate that ChatGPT would add new citations or alter existing ones while “editing.” Because the AI’s changes didn’t disturb the underlying legal propositions (which remained correct), nothing about the output looked wrong on its face.
- No second verification pass. This is the core failure: neither Gottlieb nor the law firm colleagues who reviewed the documents before filing re-ran citation checks after the ChatGPT edit. The court stressed that a competent verification process has to catch changes introduced at any stage of drafting, not just the first draft.
- Discovery came from opposing/amici parties, not internal review. The errors surfaced only after amici curiae (represented by the Jerome N. Frank Legal Services Organization) flagged the fabricated citations—not through the firm’s own safeguards.
Legal basis: The court held this violated Rule 1.1 (competence) of the Connecticut Rules of Professional Conduct, citing a growing body of authority (McCarthy v. DEA (3d Cir. 2026), Park v. Kim (2d Cir. 2024), State v. Coleman(Ohio App. 2026), among others) holding that filing AI-hallucinated citations breaches the duty of competent representation regardless of intent to deceive.
Mitigating factors: No dishonest/selfish motive, full cooperation with the disciplinary process, candid admission, sincere contrition, no prior discipline, and prompt filing of errata sheets once notified.
Sanctions imposed:
- 6 additional CLE hours in ethics/law office management (3 specifically on generative AI use), on top of the standard 12-hour annual requirement
- $1,000 donation each (Gottlieb and the firm) to the CT Bar Institute
- A compliance report to the Appellate Clerk within six months, with a courtesy copy to the Statewide Grievance Committee (not a referral)
- Compliance with reciprocal discipline in any other jurisdiction where he’s admitted
The court also used the opinion to flag Connecticut’s newly adopted practice rules (Practice Book §§ 4-9(b), 4-2(b), 62-6(d), 85-2(11)) requiring independent verification of any AI-generated citations or authorities in court filings, and noted—without deciding—that competence in this area also implicates privilege, confidentiality, and work-product concerns when using public AI platforms.
Comment: here the lawyer handled the disciplinary process admirably.








