Trust centerHow we protect your data
Your data stays home.
Built for regulated, sensitive and sovereign environments.
DeploymentWhat leaves your environment, by option
Your boundary, your choice.
On-premise, cloud managed by AIx2, or sovereign.
What leaves in each: nothing on-premise; in our cloud, only masked text reaches model providers, never used for their training.
AccessHow we connect to your systems
Read-only, everywhere.
- A dedicated read-only account, a data contract and a named owner for every source
- No inbound connections
- We work through your jump host, with recorded sessions
- All access removed at the end
PrivacyHow personal data is handled
Personal data is masked before analysis.
- E-mail metadata only, unless you approve more
- Meeting audio deleted once transcribed
- Recording and vision only with consent, under GDPR and KVKK
PeopleHow results are reported
No one is scored.
- Results by team, line and shift, never by individual
- Agreed with your works council where there is one
Human oversightWho approves what, and when
People decide.
- A named owner approves every finding and every model update
- A second person approves before anything is published
- Model updates arrive as signed packages through your change control
AI governanceHow AI is kept in check
Computed by code. Explained by AI.
- Every figure produced by code and traceable
- Confidence scores on every AI step
- Uncertain items go to people
- Support for EU AI Act transparency and human-oversight duties
Atlas securityThe safeguards around the platform itself
How we protect Atlas itself.
Role-based access
Encryption
A tamper-evident audit log
A software bill of materials
Penetration testing
Placeholder until OQ-15 is answered
Incident notice within 24 hours
Placeholder until OQ-16 is answered
Deletion at exit confirmed in writing
RegulationsThe rules our controls are designed around
Designed to support your obligations.
- GDPR
- EU AI Act
- NIS2
- Saudi PDPL and NDMO standards
- KVKK
- Quebec Law 25
- PIPEDA