Where Security Fits in an AI Agent Stack
NVIDIA, Friday, August 21st, 2026
NVIDIA's AI safety and security teams map where security controls belong as agents take on longer, more complex work.
NVIDIA's AI safety and security teams examine where security controls belong within an AI agent stack. As agents become more capable and operate over longer horizons, building security and trust into the applications they power becomes increasingly important.
The post breaks the stack into layers and identifies which risks are best handled at each. It argues that placing controls at the wrong layer produces either gaps or unusable friction. Authors are Johnny Greco, Kirit Thadaka, Ali Golshan and Alex Watson.