Netanyahu Advisor's AI Tool Exposes Security Flaw

Jonatan Urich, a long-time advisor to Israeli Prime Minister Benjamin Netanyahu, developed an artificial intelligence tool designed to continuously monitor news related to the Prime Minister, his wife Sara Netanyahu, and other Israeli political figures. The system, built on Anthropic's Claude model, automatically analyzed approximately 50 news sources every 90 seconds. Relevant content was then summarized and sent to two WhatsApp groups: one for Benjamin Netanyahu and the Likud party, and another specifically for Sara Netanyahu.

The tool also tracked publications concerning key political opponents, including Gadi Eisenkot, Yaïr Golan, Naftali Bennett, Yaïr Lapid, and Avigdor Liberman, as well as major Israeli parties and election polls. The program assigned an importance level to each piece of information and could suggest media responses for sensitive topics.

However, the system's code was published on GitHub with insufficient security, potentially exposing private phone numbers of senior officials and a key that could grant access to WhatsApp group data and allow messages to be sent via the bot. Access to the program was restricted after its existence and vulnerabilities were reported to Urich.

Reports indicate the system had been operational since at least September 1st and had already performed over 19,000 media content analyses. Urich is reportedly involved in separate legal proceedings unrelated to this security breach.