AI ETHICS
3.
Rule 3.3 – Candor Toward the Tribunal
Model Rule 3.3 prohibits lawyers from making false statements of fact or law to a tribunal and requires lawyers to correct false statements previously made. This duty of candor applies with full force to AI-assisted work product.
The most prominent AI-related candor failures have involved fabricated legal citations or propositions later noticed by a judge or opposing counsel. Human verification against authoritative sources remains essential. Think of this intervention as a “lawyer-in-the-loop,” a version of “human-in-the-loop” oversight for ensuring AI outputs are accurate. Many courts have responded to AI-related candor concerns by requiring disclosure of AI use. Federal courts in nearly two dozen states now have standing orders requiring disclosure and/or verification of AI use. Some simply require disclosure of its use in filings or other documents, but others, like Judge Brantley Starr of the Northern District of Texas, have adopted standing orders requiring attorneys to certify that no portion of any filing was drafted by generative AI without human review, or that any AI-generated content was checked for accuracy by a human. Defenders must distinguish between AI outputs and their own human analysis. When AI generates a draft, the defender must review that draft, verify any factual or legal assertions, and take responsibility for the final work product (and take responsibility for the work product of junior defenders or staff who may have used AI, too). If AI contributes to a filing, defenders should be prepared to disclose that contribution if required by court order or if candor otherwise demands it. AI can generate convincing fake audio, video, and images — so-called “deepfakes” that may be difficult or impossible to distinguish from authentic recordings. This capability creates risks that fabricated evidence might enter criminal proceedings, whether through prosecutorial misconduct, witness fabrication, or third-party manipulation. Defenders must approach audiovisual evidence with heightened scrutiny. When recordings seem inconsistent with other evidence, when provenance is unclear, or when the evidence seems too convenient, defenders should consider whether manipulation might explain the anomalies. Expert analysis may be necessary to detect sophisticated deepfakes. Defenders should also ensure their own evidence is authentic. When clients or witnesses provide recordings, letters, or other materials, defenders should verify authenticity before relying on them. Submitting fabricated evidence — even unknowingly — can devastate a client’s case and the defender’s credibility.
Beyond deepfakes, AI raises broader questions about evidence authentication. Traditional authentication requires testimony that evidence is what it purports to be. When AI generates or processes evidence,
29
Parity in Practice: The Defender’s Duty to Ethically Use AI
Made with FlippingBook Online document maker