Trust
Security & Responsible AI
How we handle your data and keep people in control. The binding terms for each engagement are in your signed agreement.
Draft: to be reviewed.
Your data
Where possible, systems run in your own cloud account. Otherwise they run on named sub-processors in a region you choose. Your data is never used to train models. Retention and deletion are set out in the contract.
Access
We use least-privilege access, 2-step verification on all company accounts, and access logs.
AI models
We choose the model provider with you for each engagement, and use only providers whose enterprise terms exclude training on client data.
People in the loop
Actions with real impact, such as payments, HR decisions or commitments to customers, need a person's approval.
Incidents
If something goes wrong, we tell you promptly. Where the law applies, we report incidents to CERT-In within the required time.
Responsible AI
- Every system is tested against an agreed set of cases (evaluations) before launch.
- No fully automated decisions about people without human review.
- We are open about what the AI does and where it can be wrong.
Questions: info@agentmesh.co.in