Netanyahu Aide's AI Tool Leaks Sensitive Data

A senior advisor to Israeli Prime Minister Benjamin Netanyahu, Jonathan Urich, developed an artificial intelligence tool designed to monitor media coverage of the prime minister and his wife, Sarah Netanyahu. The tool, built using Anthropic's Claude technology, reportedly scanned approximately 50 news sources every 90 seconds and sent summaries to WhatsApp groups. One group was dedicated to news about the prime minister and his Likud party, while another focused on mentions of Sarah Netanyahu. The system also tracked social media posts from journalists and popular Telegram channels.

The AI tool was uploaded to the code-sharing site GitHub, where it remained unencrypted and publicly accessible. This oversight led to the exposure of personal phone numbers belonging to senior officials and vulnerable access tokens. Haaretz reported the security breach, prompting Urich to restrict access to the account and remove the tool after being contacted by the newspaper. The system had been operational continuously, performing at least 19,000 scans.

Beyond monitoring mentions of the Netanyahus, the program tracked political rivals such as Gadi Eisenkot, Yair Golan, Naftali Bennett, Yair Lapid, and Avigdor Lieberman, as well as leading political parties. It assigned weighted scores to reports and compiled summaries of public opinion polls. Urich also programmed the tool to generate suggested media responses to each report, though a central component of the program was reportedly malfunctioning.

This incident adds to existing scrutiny of Urich, who is involved in several high-profile investigations. He was previously accused of transmitting classified information to harm state security in relation to a secret document leak to the German newspaper Bild. He is also a suspect in a case concerning funds received for a public relations campaign supporting Qatar. These controversies have drawn significant criticism towards the prime minister's inner circle.