Hackers Exploit Chatbots Using New Techniques

Reports reveal that hacking groups are now turning to develop more sophisticated methods to breach chatbots and artificial intelligence systems, by psychologically manipulating commands and context or exploiting hidden audio signals. This allows them to steal sensitive data or gain control over systems without detection. Previously, attackers succeeded in forcing models to provide dangerous information such as bomb-making instructions or malicious software. Additionally, exploiting psychological vulnerabilities in modern models has become a primary tactic for attacks. The study highlights that adversarial audio technology has enabled attackers to deceive voice assistant systems into executing covert commands, increasingly endangering personal and private data.

Size: 120%Text Size

Notice: This Is an AI-Generated Summary

Display The Full Article