Agentic AI expands attack surfaces by ingesting untrusted context from external platforms, emails, and memory. This session will reveal two critical exploits in major tech companies where hidden instructions hijacked agents to exfiltrate sensitive data and execute code. Attendees will see live demos and leave with concrete defenses for securely deploying agentic AI.
Additional Information: This session was developed collaboratively with Sasi Levi, Security Research Lead, Noma Security.