Configured autonomy level
Simulated demo environment · autonomy is never assumed to be infallible
Level 4 — Autonomous defense
Petavora may execute predefined containment actions on confirmed threats, within policy.
Human role
Human is notified immediately and holds a rollback control.
Petavora may
- Isolate a workload
- Restrict an AI agent
- Rotate an exposed credential
- Block an east-west flow
Petavora may not
- Destructive actions
- Anything outside the named containment set
Executed automatically
05
Awaiting human decision
03
Blocked by policy
01
Rolled back
01
Action risk classes
Reversible, no user-visible disruption expected.
2 of 2 automatic at Level 4
Limited, scoped disruption; reversible with a known procedure.
3 of 4 automatic at Level 4
Material disruption to people or services; needs a named approver.
0 of 4 automatic at Level 4
Data or capacity loss possible. Never automated under any level.
0 of 1 automatic at Level 4
What each level unlocks
L0 · Observe
Monitor only. Petavora records and correlates, and takes no action of any kind.
EnabledL1 · Recommend
Petavora proposes defensive actions with evidence; a human decides and executes.
EnabledL2 · Investigate
Petavora automatically gathers context — read-only enrichment, no change to the environment.
EnabledL3 · Controlled defense
Petavora may execute approved low-risk (SAFE) actions inside named policies.
EnabledL4 · Autonomous defense
Petavora may execute predefined containment actions on confirmed threats, within policy.
EnabledL5 · Predictive defense
Petavora may act on high-confidence predicted attack paths before the step occurs — only where explicitly authorised, per scope.
OffAutonomous defense is not infallible. Confidence values are model estimates, simulation results are expected outcomes rather than guarantees, and predicted attack steps may never occur. Every automated decision is recorded with its evidence and policy, and is reversible wherever a rollback path exists. All data on this screen is simulated demo telemetry.