Israel Cyber Chief Warns Iran is Developing AI-Powered Cyberattacks
National Cyber Directorate head Yossi Karadi warned that Iran is developing AI-powered cyberattack capabilities. Speaking at Cyber Week, he highlighted the emerging threat of a cyber war of attrition and raised concerns over AI-driven physical risks.
Head of Israel's National Cyber Directorate Yossi Karadi warned on Monday at the Cyber Week conference that Iran is developing AI-driven cyberattack capabilities that could threaten the entire globe. "While the world works to prevent Iran from acquiring nuclear weapons, we know that Iran is also working to develop additional offensive cyber capabilities utilizing artificial intelligence," Karadi stated in his address.
Karadi outlined the temporal gap where attackers hold an advantage over defenders before AI capabilities are fully integrated into national defense systems. He noted that an attacker needs to use a single model, target a single objective, and choose when to strike, whereas defenders must simultaneously protect numerous organizations around the clock against attacks powered by a wide range of models while complying with regulatory requirements.
"At the national level, the task of defenders is much harder," Karadi said. "We must simultaneously and successfully protect numerous organizations, around the clock, against attacks enabled by a wide array of different models, all while meeting regulatory demands," he added.
Later in his speech, Karadi presented a new concept of cyber war of attrition, recently published in an American journal article. According to this framework, the strategic threat lies not merely in the power of a single strike, but in the accumulation of attacks over time, where the frequency of assaults outpaces recovery rates, potentially leading to systemic national damage from cumulative localized incidents.
"In the new cyber war of attrition, the enemy does not need to win in a single attack," Karadi explained. "They can attempt to achieve victory through continuous attrition. A state that fails to recover at the pace it is attacked could find itself gradually worn down, even without a single decisive cyberattack," the directorate head said.
Concluding his address, Karadi hosted Roberta, a humanoid robot from Prof. Yuval Elovici's laboratory, on stage, raising questions about human control over machines. "When a machine operates in the physical world, are we still in control?" he asked. "Imagine a million robots from the same manufacturer operating across dozens of countries. Then, one day, an update or a cyberattack causes all million robots to execute the exact same dangerous action at the exact same time. People could be harmed."
