Zero-Training Data Retention
Your proprietary agent logic, terminal traces, and tool payloads are strictly isolated. We never use your execution data, logs, or payload contents to train our internal models or any third-party AI models. Reality Kernel is security infrastructure, not an AI data broker.
What we collect
We collect only the data needed to enforce cryptographic boundaries and provide auditability:
- Verdict Logs: We store cryptographically hashed signatures of the commands your agent attempts to run, mapped to an ALLOW/BLOCK status. Sensitive payload contents are redacted or irreversibly hashed before touching our audit ledger.
- Tenant Configuration: Multi-tenant identifiers, API quotas, policy definitions, and access logs.
- Operational Telemetry: Standard performance metrics (latency, error rates) to maintain our 0.31ms p50 SLA.
Contact & Compliance
For data deletion requests, GDPR/CCPA inquiries, or security disclosures, contact our compliance team at contact@realitykernel.dev.