Israel Uses AI System to Target Hezbollah Figures in Lebanon

Even SIM card changes would not prevent tracking, according to an AI specialist who previously worked with defense firms. The specialist, granted anonymity, described a data pipeline that includes phone metadata, location pings, app usage, social media behavior, banking records and facial recognition inputs.

Much of the information is scraped from commercial platforms, mobile networks and partner agencies. Platforms such as Palantir’s Maven standardize, tag and score the data, linking it across devices and accounts. Palantir has spoken openly about its work with the Israeli military.

The AI then builds timelines of activity and maps networks of relationships.

Mounir Shehadeh, who served as Lebanon’s coordinator with the United Nations peacekeeping mission, said Israel has had access to much of Lebanon’s data infrastructure for two decades. This includes mobile subscriber and vehicle registration databases.

Israel also hacked into Hezbollah’s terrestrial network and signal corps, he said. Hezbollah’s involvement in the Syrian civil war from 2011 to 2024 further compromised its security. These factors enabled Israel to build a target list covering both field commanders and senior leaders, Shehadeh said.

The AI processes terabytes of data to detect patterns, compare movements to known threats and analyze deviations from routines. This creates a threat profile. Col. Yoav, head of the Israeli military’s Artificial Intelligence Center, said in a February 2023 article that the system completes the process in seconds, compared with weeks required by teams of investigators in the past.

The AI specialist raised concerns that the systems rely on data correlations rather than verified logic. Flawed information can lead to repeated errors with greater speed and confidence. The specialist questioned whether decisions are made by humans or algorithms and how the system distinguishes identities.

Vasji Badalic, a professor at the Institute of Criminology in Slovenia who wrote a 2023 paper on metadata-driven targeting, said such systems can produce false positives. They may classify relatives, propaganda workers or financial coordinators as combatants based on communication patterns.

The threshold separating combatants from civilians remains unclear. The approach is not new. The U.S. military used phone metadata analysis during the Iraq war and the National Security Agency ran a behavioral profiling program called SKYNET in Afghanistan.