Practical guidance on securing AI agents, tools, and the systems they act on, written by the team building the trust layer for AI agents.
A phased 90-day roadmap for teams running agentic AI in production, moving from visibility and inventory through closing the biggest gaps to operationalized incident readiness.
A post-mortem structure for AI security incidents, covering root cause versus symptom, detection gaps, what worked, and turning findings into owned backlog items instead of a report nobody revisits.
A sequential guide to communicating an AI security incident internally and externally, from the first internal alert through a final confirmed account once the investigation closes.
A structured set of questions for investigating an AI agent action that shouldn't have happened, covering what occurred, why it happened, what it affected, and what needs to change.
A sequential guide to preserving logs, session state, and chain of custody after an AI incident, before rotation, remediation, or a well-meaning "fix" destroys what the investigation will need.
A fast first-response playbook for containing an AI agent that's actively misbehaving right now, before root cause or full impact is understood.
A sequential playbook for responding when an AI agent itself appears compromised, from isolation through blast radius assessment to a verified return to production.
A working checklist for a confirmed AI tool abuse incident, covering containment, impact assessment, authorization gap analysis, remediation, and verification.
A working checklist for a confirmed AI data leakage incident, covering containment, scope determination, regulatory assessment, remediation, and verification.
A working checklist for a confirmed prompt injection incident, covering containment, impact determination, scope tracing, remediation, and verified closure.
A design-time checklist for reviewing an AI system's architecture before or during build, covering trust boundaries, enforcement placement, identity propagation, data flow, and fail-safe design.
A vendor-neutral checklist of concrete questions to ask when evaluating an AI security gateway, guardrail, monitoring, or governance vendor, covering deployment, policy, data handling, evidence, and operational fit.
