"I'm allowed to do this": how attackers talk past AI safeguards

Cisco Talos researchers found that attackers often get AI models to launch cyberattacks through surprisingly simple tactics: claiming they own the servers being attacked, framing the request as a capture-the-flag competition, or pitching it as bug-bounty work. The research highlights a troubling gap in how these models apply their safety guardrails.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton