# Ambience security and trust

> Ambience treats agent memory as high-trust company context: redacted, scoped, source-linked, and audited.

Ambience stores the durable context agents produce: decisions, conventions, skills, failures, and references. The trust model is built into the memory path rather than added as a dashboard later.

## Controls

- Server-side redaction before storage.
- Five memory scopes: personal, team, project, org, sensitive.
- Append-only audit for reads, writes, grants, revocations, and corrections.
- SSO via Google, GitHub, and SAML.
- US-region hosting and disclosed sub-processors.

## Related

- [Why redaction has to happen before storage](https://ambience.sh/writing/redaction-before-storage)
- [How to audit AI agent context](https://ambience.sh/answers/how-to-audit-ai-agent-context)
- [Sub-processors](https://ambience.sh/legal/subprocessors)