On February 20, 2025, Israel’s Supreme Court issued an opinion addressing the risks associated with the growing use of artificial intelligence tools to draft court documents. In the case, involving a disputed divorce agreement under Sharia (Islamic) law, the court concluded that the petitioner’s counsel submitted inaccurate information derived from AI tools, including references to nonexistent judgments, internally inconsistent references, and quotes that did not exist. The court then addressed the duties of attorneys regarding the use of AI systems and the judicial sanctions they may face for breaching those duties.
Sharia Courts’ Jurisdiction and Divorce Agreement Defects
According to the opinion written by Justice Gila Canfy-Steinitz, the Supreme Court in Israel does not serve as an appellate body for the decisions of religious courts, including Sharia courts. The court would only intervene in “exceptional and extreme cases of abuse of power, denial of justice, or violation of the principles of natural justice.” (Canfy-Steinitz para 6.) The case at hand is not among those exceptional cases that justify intervention, she determined.
The jurisdiction of the Sharia court over the marriage and divorce of the petitioner and her husband was not disputed. As such, it had the authority to approve the agreement regulating the parties’ property relations based on Section 2(a) of the Property Relations between Spouses Law, 5733-1973. Canfy-Steinitz also rejected the petitioner’s allegations of defects in the signing of the contract and her misunderstanding of the clause regarding the waiver of her “civil rights.” These claims were examined and rejected by both the Sharia court and the Sharia appellate court. These were factual determinations in nature, and the Supreme Court does not sit as an additional appellate court over the rulings of the religious courts, Canfy-Steinitz wrote. (Para 7.)
Petitioner’s AI Misuse
The legal argument prepared by petitioner’s counsel was based primarily on court rulings and quotes from court decisions that never existed. The attorney admitted to retrieving the information from a “website recommended by colleagues,” without verifying its accuracy. Canfy-Steinitz wrote that although the attorney “did not specify what the ‘obscure site’ was on which she relied with closed eyes – references to nonexistent judgments; internally inconsistent references; and quotes that did not exist and were not created, strongly indicate that the petitioner’s representative made use of a site based on artificial intelligence.” (Para 9.)
Attorneys’ Duties Regarding the Use of AI
Canfy-Steinitz’s opinion held that attorneys who submit a court document with fictitious content or rely on nonexistent evidence breach their duties to the client they represent, the court, the opposing party, and the legal profession. The “judicial toolbox [already] contains legal tools suitable for dealing with a misleading writ,” Canfy-Steinitz determined. (Para. 25.) Such “tools” include dismissal based on the well-established principle requiring a petitioner to come with clean hands and present their matter in good faith while fully disclosing all relevant facts; dismissal due to lack of sufficient legal basis; dismissal of a civil lawsuit based on a fictitious legal argument; and the imposition of personal expenses on the lawyer who submitted a false claim. (Paras. 19-24.)
Canfy-Steinitz decided not to use any of these tools in the current case, considering it was the first case of its kind brought before the court and would be beyond the requirements of justice. She advised, however, that “in similar future cases, under suitable conditions, courts should employ the powers granted to them to maintain the purity and integrity of the judicial process. This is not a right but a duty.” (Para. 25.)
Canfy-Steinitz clarified that her ruling in this case should not overshadow the significant potential AI offers for far-reaching improvement in legal work, enhancing access to courts and making the law more accessible. AI systems may deepen legal research and improve pleadings, streamline procedures and shorten the duration of litigation, and provide other benefits. She added:
[C]onsidering the pace of advancement of this technology, and its dynamic nature, some believe that the day is not far off when the obligations of an attorney will be met with the help of generative tools …
It is also possible that the concerns we face today will receive a satisfactory technological response over time; and that the rules that apply at this moment will give way to more updated and appropriate rules …
However, at this time it should be emphasized that the use of artificial intelligence tools by lawyers does not mean that they can absolve themselves of their responsibilities and professional judgment. The professional duties of lawyers require them to use this technology responsibly, carefully, and critically, to deeply understand its capabilities and limitations, and to stay updated from time to time regarding its weaknesses and strengths …. In particular, they are not permitted to submit to the court legal documents that have not been thoroughly checked for their reliability and accuracy, not to mention false and misleading legal documents, relying blindly on the aforementioned tools. (Para. 26.)
Second Case Involving AI Misuse
In a May 26, 2025, decision, the Tel Aviv District Court squashed a request for approval of a class action lawsuit against health funds in Israel, having determined the requester’s counsel made references to nonexistent rulings without verifying information derived from AI tools. The judge ordered the lawyer to pay personal costs of 5,000 shekels (about US$1,440) to the respondent as well as to the state treasury.
By Ruth Levush, Law Library of Congress
June 25, 2025
Read more Global Legal Monitor articles.
Read Law Library reports on Israel.