Have You Been Using AI Without Verifying? It Could Cost You ILS 30,000

Israel’s Supreme Court is stepping up its efforts to address AI-generated hallucinations, and recently ordered the Ramat Gan Municipality to pay exceptional court expenses of ILS 30,000. This is due to the municipality’s reliance on fictitious case law and a non-existent Ministry of Education Director-General circular, in a way that violated a special education child’s right to transportation to his school.

AI in Pleadings: Permitted, But Don’t Get Carried Away

Artificial intelligence use has long become routine, including in the legal world. But there is a huge difference between thoughtful, responsible use of AI—as a tool to help produce results that are then scrutinized and verified by human discretion—and unchecked use, whereby AI-generated text is included in pleadings or official responses without any filtering or verification.

This is especially true when the user bears responsibility for presenting an accurate factual and legal picture – what the law says and what has actually been determined in case law.

Israeli Courts’ Attitude Toward AI Hallucinations in Pleadings

Over the past year, several court rulings have sharply criticized a disturbing phenomenon: erroneous references, citations that never existed, and fictitious legal provisions, all resulting from the use of AI tools without independent professional verification. Until now, the courts have imposed sanctions ranging from a few hundred to a few thousand shekels on litigants who submitted unverified AI-generated material to the court.

The Supreme Court’s recent ruling signals that judicial patience is running out. This is translating into heavier sanctions and particularly high standards of conduct for public authorities.

Cohen v. Ramat Gan Municipality: Not a “Technical Error” But a Cumulative Failure

The Supreme Court’s ruling in the Ramat Gan Municipality case paints an unusually severe picture. In its administrative response to a person seeking to exercise his special needs child’s right to transportation to school, the municipality relied on “Ministry of Education Director-General Circular 5783/4(a),” which does not exist, in order to deprive the child of his right to transportation. Not only that, but the municipality cited detailed sections from the fictitious circular that do not appear in any official procedure.   

Later, during the legal proceedings, the municipality referred to case law that ostensibly supported its position, but, in reality, the cited rulings did not contain the alleged content, and some references did not exist at all.

The Supreme Court sharply criticized the municipality’s conduct. The focus was not only on the existence of errors, but also on the context – a municipal authority who is responding to a citizen and managing a legal proceeding from a position of power.

The court highlighted the power imbalance and noted that ordinary citizens typically lack the ability to identify arguments that may sound legally persuasive but are, in fact, entirely fake. While lawyers’ pleadings are refutable by opposing counsel and the court itself, direct interactions with public authorities carry far greater potential harm. Vulnerable members of the public may be harmed without realizing that something is wrong. Therefore, the Supreme Court ruled that the municipality’s unverified use of AI-generated content constituted reckless and unlawful conduct.

The Supreme Court did not make do with scathing criticism; it ordered the municipality to pay ILS 30,000 in expenses. This sends a clear and unequivocal message: the era of judicial patience is over. Every entity must ensure its use of AI is correct, controlled, and responsible, and may not shirk its duty to exercise accountability, judgment, and oversight over all material it produces.

Israeli Courts Are Not Banning AI, But Demanding Full Human Accountability

This Supreme Court ruling marks a new phase: the judiciary will no longer make do with warnings or symbolic expenses. It is prepared to impose heavy sanctions, especially when unchecked AI use undermines the integrity of legal proceedings, misleads the courts and opposing parties, and risks citizens blindly accepting administrative decisions as if they were “based on law.”

The practical conclusion for lawyers, authorities, and anyone submitting pleadings or making administrative decisions in the spirit of “AI-first” is simple: you may use AI tools, but don’t blindly believe AI-generated output. Independent verification is not optional but a fundamental prerequisite. Courts are emphatically demonstrating that they have lost all patience with a phenomenon that threatens both the integrity of the judicial process and public trust in authorities and the judicial system.

It is important to note that the Supreme Court is not seeking to ban AI use and even acknowledges its possible benefits. But the message is unequivocal: AI is a work tool, not a substitute for performing your job. If you use AI, you must verify output and sources and actually exercise human discretion.

[Adm. Appeal 63194-08-25 Cohen v. Ramat Gan Municipality, 22.3.2026.]

***

Adv. Gal Livshits is a partner in the firm’s Litigation Department. 

Barnea’s Litigation Department is one of Israel’s leading practices, ranked in the top tiers by both local and international legal directories. We represent some of the country’s most prominent clients in complex legal disputes, both local and international. Our firm has extensive experience in representing leading clients in the Israeli market, including public and private companies, government entities and local authorities, banks and financial institutions, and directors and senior executives.