Autonomy that can survive being asked hard questions afterward.
This page is written for the conversation that follows an incident, not just the one before it — including the parts of the governance model that are still drafts.
Generated continuously, not assembled for an audit.
Predator's Governance & Risk Intelligence layer (L5) maintains mapping against the following frameworks as a byproduct of normal operation.
Three rules that hold regardless of confidence score.
Hard guardrails, not soft preferences
Certain actions — deleting production databases, modifying billing configuration — are never permitted autonomously. This is enforced architecturally at L6 before any action reaches the Remediation Fabric, not configured as a tunable preference.
Evidence before action
No autonomous action executes without a logged evidence trail and a plain-language reasoning record. The console demo on this site shows this exact record for every incident, autonomous or escalated.
Confidence has a ceiling on authority
A high confidence score expands what Predator may do autonomously within its governance tier — it never removes the human approval requirement for actions outside that tier, regardless of how confident the system is.
Who answers for what.
Actual liability for autonomous-system decisions depends on jurisdiction, contract terms, and the specific facts of an incident. The allocation below reflects Predator Defense Systems' reviewed and approved position on how responsibility is assigned across the layers of the platform.
Built, deployed, and running — not conceptual.
Every layer described on this site reflects production architecture.
Everything described in Platform Architecture and demonstrated in the console demo reflects the platform as currently built.
Have a compliance or legal question we didn't cover?
We'd rather answer it directly than have you guess.