Sanctions Sought Over Alleged AI-Generated Fake Citations in Connecticut Case
A motion for sanctions in a Greenwich-based breach of contract case alleges the defense used AI-assisted research that produced fake case law. The filings reportedly included quotes that don't exist and citations to cases that can't be found. The plaintiff's counsel argues this crosses ethical lines and undermines the court's process.
The dispute stems from a December 2024 lawsuit by J. Salvatore & Sons Inc. against Gencor Contracting Corp., claiming defective masonry led to flooding and more than $80,000 in damage at a Midwood Road project in Greenwich. The sanctions motion, filed Jan. 16 in state Superior Court in Danbury, targets two motions to strike submitted by defense counsel in October and November.
What the Motion Alleges
The first motion to strike allegedly contained 12 misrepresentations, including fabricated quotes and non-existent cases. After the plaintiff's team flagged suspected AI-generated content on Nov. 3, defense counsel acknowledged their research software "has an AI component" and withdrew the motion, promising to refile with different citations.
The second motion was filed-but the plaintiff says it "also suffers from AI fabrications," including the same fake quote. "The unfettered use of AI without ensuring the accuracy of the content submitted to the Court is problematic," the motion states, calling it a violation of duties of accuracy, candor, and professionalism. The defense attorney could not be reached for comment Tuesday.
Why This Matters for Practitioners
Connecticut's Rules of Professional Conduct prohibit false statements of material fact or law to a tribunal. The motion notes a shortage of direct authority on AI misuse in court filings, but the principle is familiar: you own the accuracy of what you file. AI tools don't change that duty.
The Rules Committee of the Connecticut Superior Court discussed a proposal on Jan. 12 to require lawyers who use AI for legal research to certify that they verified citations. No action was taken, but the signal is clear-verification is on the record. Federal judges are already warning counsel about AI hallucinations.
Bench Signals You Should Note
In September, the U.S. District Court in Connecticut issued a notice cautioning lawyers to verify AI-assisted content "like any other shoddy research method." It stated a no-tolerance policy for briefing-AI-assisted or not-that hallucinates legal propositions or badly misstates the law. Expect more explicit requirements and certifications to follow in various courts.
The American Bar Association's Task Force on Law and Artificial Intelligence released guidance in December emphasizing competence, integrity, and public trust. "The future of our profession will be shaped by how we meet this moment," wrote the task force's immediate past president. You can review the task force materials here: ABA Law & AI.
For developments at the state level, track agendas and minutes here: Connecticut Rules Committee.
What Sanctions Are Sought
The motion seeks reimbursement for the time spent investigating and addressing the alleged AI hallucinations in both motions to strike, plus any further relief the court finds appropriate. Sanctions aside, the reputational cost is real. Courts have long memories of filings that waste time.
Practical Safeguards for Using AI in Legal Research
- Run every citation through citators and pull the full opinion. Confirm quotes word-for-word and ensure the proposition matches the holding.
- Cross-check with primary sources (official reporters, court websites, and dockets). If you can't find it in a trusted database, don't use it.
- Require internal verification logs. Document who checked what, when, and where.
- Disable or limit generative features in drafting tools unless a human validator reviews every authority.
- Add a pre-filing citation audit step to your workflow. Make it mandatory for any AI-assisted draft.
- Monitor court-specific policies on AI usage. Some judges require certifications or disclosures.
- Train your team on AI limitations, prompt hygiene, and verification protocols. Tools are helpful; unchecked outputs are a liability.
Case Takeaway
AI can accelerate research, but it can't replace professional judgment. If you file it, you own it. Verify everything, document your process, and stay current on court expectations.
If your firm is building AI policies and training, see structured options here: AI courses by job function.
Your membership also unlocks:
