lethal trifecta

A security framework identifying three components that create vulnerabilities in AI agents: access to private information, exposure to malicious instructions, and exfiltration mechanisms.

Topics