Jerusalem, 8 October, 2026 (TPS-IL) — Israel National Cyber Directorate chief Yossi Karadi warned world leaders that artificial intelligence systems could soon make critical security decisions faster than humans can intervene, Israeli media reported Thursday.
In a strategic document sent to international leaders and cybersecurity officials, Karadi called for urgent limits on the autonomy granted to AI agents and models.
Karadi warned that increasingly automated cyberattacks will force defensive systems to detect, decide and respond at machine speed, potentially making human approval impossible for every action. Greater autonomy, however, could make AI decisions harder to understand, stop or attribute.
He cited a scenario in which an autonomous defense system responding to a zero-day attack mistakenly isolates critical infrastructure, potentially shutting down an electricity network or hospital.
Karadi proposed an “AI strategic triangle” balancing innovation, security, and human control, with irreversible national decisions designated as authority that cannot be delegated to autonomous systems.
