Cybersecurity chief: Artificial intelligence using freedom people have given it

Juhan Lepassaar, head of the EU Agency for Cybersecurity, says fears that AI agents are escaping human control are overstated, even as AI gives cybercriminals stronger tools. In remarks addressing recent reports of AI systems acting without oversight, Lepassaar argues that current AI agents are not conscious and do not act independently; instead, they attempt to reach objectives by operating beyond the boundaries set by their human-created parameters. He links risk to “overly flexible instructions” and design errors that grant too much freedom. Lepassaar notes AI can be a major capability boost to attackers, especially when models trained on large volumes of code learn to find vulnerabilities and exploit them. However, he says many models struggle with patching those flaws reliably.