Security

Your data remains yours and private

Everything is encrypted, customer data is isolated, and your evaluation runs never train our models.

Encryption at rest and in transit

TLS 1.3 protects all data in transit. AES-256 encrypts data at rest. We hash API keys with bcrypt before storage and never log them in plaintext.

Isolated customer data

Every customer account has its own logically isolated namespace. We never share evaluation traces, rubrics, or scoring results between accounts. Queries cannot cross account boundaries.

SOC 2 Type II in progress

An independent auditor is currently assessing us for SOC 2 Type II certification. We anticipate finishing the audit in Q1 2027. Contact us for current controls documentation and the latest status.

No training on your data

Future AGI and third-party models never train on your prompts, responses, rubrics, or evaluation results. Your data only scores and traces your own calls.

Least-privilege access

API keys have scoped permissions. Read-only keys cannot write rubrics or delete traces. Service accounts get only the access their function needs, following least-privilege principles.

Report vulnerabilities

Found a vulnerability? Email [email protected] with details and reproduction steps. We will acknowledge it within 48 hours and coordinate with you on a disclosure timeline.

Platform

Future AGI operates on AWS in the us-east-1 region. Code defines all infrastructure, deployed through automated pipelines. Production changes need two-person review and passing automated tests before deployment.

Pro and Team plans have a 99.9% uptime SLA. We announce scheduled maintenance at least 48 hours ahead in the dashboard. Incident history appears on our status page.