Launch

• Your data stays where you choose

  Client data stays in the client's chosen environment and region, and is never used to train public models.

• Personal data redacted first

  Personal data is detected and redacted before it reaches a model.

• Role-based access

  Staff see only what their role allows, in the AI as in your other systems.

• Human approval for high-risk actions

  Payments, refunds, legal or medical replies and deletions wait for a person to approve them.

• Prompt-injection and data-leak testing

  Every build is tested for prompt injection and data leaks before launch.

• Evaluation signed off by a data scientist

  Evaluation sets and accuracy reporting are reviewed and signed off by a data scientist.

• Full audit logs

  Logs record what the AI saw, said and did.

• Kill switch and rollback

  Any AI feature can be switched off at once and rolled back to the last good version.

• Aligned with the rules that apply to you

  Designed to align with GDPR, the Kenya Data Protection Act, the EU AI Act and sector rules as relevant.

How we secure the AI Core →