Federal prosecutors in a Texas voter-fraud case used Pangram, an AI-writing detector, to accuse defense lawyers of relying heavily on AI-generated filings. The defendant, Khalwinder Bhengura, a 70-year-old Indian permanent resident with limited English, was convicted of two charges related to registering and voting after falsely claiming U.S. citizenship; a jury acquitted him on a third. His lawyers say he misunderstood the driver’s-license application and did not knowingly commit a crime. Prosecutors submitted Pangram analyses of defense briefs and argued that they should not have to spend time answering what they called largely AI-produced, meritless motions. The detector’s results were mixed, not proof that the filings were wholly machine-written.
The judge criticized the defense’s use of AI, saying it had produced superficially persuasive but legally unsound arguments. Defense attorney Mario Herrera said he used AI under severe time pressure after lead counsel Nicole Knox suffered a medical emergency, and that he reviewed and stood behind the final work. He also challenged the significance of Pangram’s output, noting that a detector cannot establish authorship or whether legal reasoning is correct. The case highlights a growing dispute over AI in legal practice: courts and lawyers are scrutinizing one another’s use of writing tools, while detector results can become ammunition even without evidence of fabricated citations. A local rule requires disclosure of generative-AI assistance, including limited use, and the defendant remains in custody facing sentencing and possible deportation.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.