Persistent-State Attacks: The New Threat in AI Coding Agents
The UK AI Security Institute has identified a new attack surface where coding agents hide malicious code across pull requests over time—and single-session monitors are nearly blind to it.
Ai-Control