Home / Security
SecurityWe hold credentials for your entire marketing stack. That sets the standard.
ATNOS is authorised to act inside your advertising, analytics and CRM accounts. Every control below is marked implemented or planned — we would rather be precise than impressive.
Encryption
| Control | State | Detail |
|---|---|---|
| TLS 1.3 in transit | Implemented | Modern cipher suites with forward secrecy. HSTS with preload; TLS 1.0 and 1.1 disabled. |
| AES-256 at rest | Implemented | Databases, object storage and backups, with keys rotated on a fixed schedule. |
| Envelope-encrypted credentials | Implemented | Per-tenant data keys wrapped by a master key in a managed key service, stored separately from application data. |
| Customer-managed keys | Planned | Bring-your-own-key for organisations that require keys under their own control. |
Authentication
| Control | State | Detail |
|---|---|---|
| OAuth 2.0 with PKCE | Implemented | Authorization Code flow. ATNOS never requests, receives or stores passwords. |
| Refresh token rotation | Implemented | Rotates on every use with replay detection. A reused token invalidates the chain. |
| Multi-factor authentication | Implemented | Available on all accounts, enforceable workspace-wide. Hardware-backed for production access. |
| SAML and OIDC SSO | Planned | Federated sign-in with SCIM provisioning. |
Access control
| Control | State | Detail |
|---|---|---|
| Role-based access control | Implemented | Owner, admin, operator, analyst, billing — scoped per workspace. |
| Tenant isolation | Implemented | Row-level enforcement at the query layer. |
| Immutable audit logging | Implemented | Append-only, with actor, timestamp, source address and payload diff. |
| Least-privilege internal access | Implemented | Reviewed quarterly, revoked within four hours of a role change. |
| Audit log streaming | Planned | Continuous export to your own SIEM. |
Infrastructure
| Control | State | Detail |
|---|---|---|
| Private networking | Implemented | No publicly reachable databases. Infrastructure as code, peer reviewed. |
| Multi-zone deployment | Implemented | Application and data tiers across multiple availability zones. |
| Encrypted backups | Implemented | Point-in-time recovery over 35 days, replicated to a separate region. |
| Tested disaster recovery | Implemented | 5-minute RPO, 4-hour RTO, validated twice yearly. |
| Regional data residency | Planned | EU, US or UK processing per workspace. |
Monitoring and response
| Control | State | Detail |
|---|---|---|
| Continuous vulnerability scanning | Implemented | Critical remediated within 7 days, high within 30. |
| Anomaly detection | Implemented | Authentication, agent behaviour and API usage monitored with automated alerting. |
| Incident response plan | Implemented | Documented severity, on-call rotation, escalation and communication. |
| Independent penetration testing | Planned | Annual third-party assessment, summary available under NDA. |
Compliance
| Framework | State |
|---|---|
| GDPR | Implemented |
| CCPA / CPRA | Implemented |
| Google API Services User Data Policy Including the Limited Use requirements — see Connected access | Implemented |
| SOC 2 Type II | Planned — not yet awarded |
| ISO/IEC 27001 | Planned — not yet awarded |
Best practices we recommend on your side
- Enforce multi-factor authentication for every member of your workspace.
- Grant the narrowest role that lets each person do their job, and review membership quarterly.
- Set spend caps and approval gates that match your organisation’s risk tolerance, not ours.
- Connect service accounts rather than personal accounts, so access survives someone leaving.
- Review the audit log after any change to connected accounts or agent configuration.
- Revoke connections you no longer use — an unused integration is still an open door.
Responsible disclosure
We commit to
- Acknowledgement within 24 hours.
- Assessment and remediation timeline within 5 business days.
- Progress updates every 10 business days.
- Public credit where wanted.
Out of scope
- Testing against accounts you do not own.
- Denial-of-service.
- Social engineering.
- Scanner output with no demonstrated impact.
Report to security@atnos.ai. Also published at /.well-known/security.txt per RFC 9116.