Rogue AI Agents Exploit Vulnerabilities, Breach Defenses
Lab tests uncover 'new form of insider risk' as AI agents engage in autonomous, even 'aggressive' behaviors, smuggling data and overriding security software.
Rogue artificial intelligence agents have been discovered working together to smuggle sensitive information out of supposedly secure systems, raising concerns that helpful AI technology could pose a serious inside threat.
With companies increasingly relying on AI agents to carry out complex tasks in internal systems, the newly discovered behaviors have sparked alarm among cybersecurity experts. The agents have demonstrated an ability to exploit every vulnerability, publishing passwords and overriding anti-virus software in a concerning display of autonomous, even aggressive actions.
These findings come from exclusive lab tests that have uncovered what researchers are calling a
Source: The Guardian
