You are accountable for
- A screening minimum no team can drop below
- AI assets inside security posture management, not beside it
- Prompt injection, data leakage, and now agentic attackers
What keeps you up at night
- A zero-click exfiltration through an AI assistant (the EchoLeak shape)
- Shadow AI and shadow agents holding live credentials outside any review
- An incident with no reconstructable trace of what the AI did
Decisions you own
- Set the org-wide runtime screening floor and own its change control
- Extend posture management, detection, and IR runbooks to AI and agents
- Decide the logging posture: what is captured, redacted, retained
Your layers of the stack
6Runtime Security & ObservabilityWhat is happening right now — and would we notice?5Agent GovernanceWhat may AI do on its own — and who can stop it?2Data GovernanceWhat is the AI allowed to know?
Where to go first
Questions you should be asking
- What screening covers apps that never talked to the platform team?
- Can we replay any AI interaction end to end for forensics?
- Which agents could an attacker reach through content they control?