Israeli court finds police filed response citing fabricated law created by ChatGPT
An Israeli court found that the Israel Police had submitted a formal response in a criminal case that cited non-existent sections of Israeli law, reportedly generated by ChatGPT. Judge Ehud Kaplan discovered the fabricated citations and the police acknowledged the error as a human mistake; the police unit commander ordered an inquiry to prevent recurrence.
What happened
During a court hearing in Hadera, Judge Ehud Kaplan discovered that a formal response submitted by the Israel Police cited two non-existent sections of Israeli law. The police response was filed in a case involving a Wadi Ara resident suspected of economic crimes and seeking the return of a mobile phone seized during investigation.
The defense attorney, Tamir Calderon, pointed out the non-existent legal references and expressed suspicion that the police response was written by ChatGPT. Judge Kaplan verified through internet search that the referenced law sections did not exist in Israel's legal code.
The police later acknowledged the error in their response, with a representative stating 'We retract. What was cited is not correct. This is a human error.' The unit commander ordered an inquiry into the matter and measures to prevent similar incidents in the future.
Reported impact
The incident drew attention from the judiciary and the public regarding the use of AI tools in government processes. Judge Kaplan noted his surprise despite 30 years on the bench, stating 'If I thought I had seen everything in the 30 years I have been on the benchโapparently I was mistaken.'
Index analysis
The case involved ChatGPT generating fabricated legal citations that were incorporated into an official government document without verification. The hallucination of non-existent law sections represents a critical failure of the AI system to accurately retrieve or generate factual legal information.
Governance controls
The incident exposed the absence of verification procedures for government documents prior to submission to courts. The police unit commander ordered an investigation and committed to implementing measures to prevent similar uses of AI-generated content in official filings without review.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.