- Revoke affected model, forge, integration, and service credentials.
- Pause impacted queues and preserve audit, provider, and deployment evidence.
- If execution isolation may be compromised, destroy the worker pool and caches before resuming.
- Rotate secret versions and restart workloads so they reload credentials.
- Restore regional service from a validated database checkpoint and matching object versions when required.
- Document impact, data classes, containment, corrective actions, and customer notification decisions.
Trust
Incident response
Containment and recovery priorities
Last modified on July 28, 2026
⌘I

