Trust and security

AI your IT director can sign off on.

Written for IT directors and data protection officers, readable by anyone. Here is how we handle your data, your access rights and control over agents.

  • Data in your environment
  • No training on your data
  • Human approval
  • Full audit log

01

Your data stays in your environment

We deploy in the cloud of your choice or on your own servers. Your data does not pass through our infrastructure.

In practice

  • Deployment in your AWS, Azure or GCP account, or on premise
  • Encryption in transit and at rest
  • No copy of data outside your perimeter

AWS · AZURE · GCP · ON PREMISE

02

No model training on your data

We use the enterprise offerings of model providers, which exclude training on your data. Your information is used only for your own processing.

In practice

  • Contracts and settings with no training and no extended retention
  • Models hosted on your side when needed
  • Sensitive data masked before processing

03

Human approval at critical steps

You decide which actions need sign off: creating an order, writing to a customer, triggering a payment. The agent prepares, a person approves.

In practice

  • Configurable approval rules and thresholds
  • Approval from Slack, Teams, email or your business tool
  • Edits possible before any action

04

Complete log of every agent action

Every read, decision and write by the agent is timestamped and kept. You always know who did what, when and why.

In practice

  • Searchable log, exportable to your SIEM
  • Sources traced for every answer
  • Retention period agreed with you

05

Access through your SSO and permissions

Agents authenticate through your identity provider and only reach what the user is allowed to see.

In practice

  • SAML and OpenID Connect
  • Permissions inherited from your existing systems
  • Least privilege service accounts

SAML · OIDC · ENTRA ID · GOOGLE · OKTA

06

GDPR compliance and data residency

Hosting in the region of your choice, data minimisation and documentation ready for your data protection officer.

In practice

  • Data residency in Europe or Morocco as required
  • Support for records of processing and impact assessments
  • Data processing agreement (DPA)

GDPR · CNDP

07

Your choice of AI models

Commercial or open source models, hosted by a provider or by you. You are never locked into a single vendor.

In practice

  • Model chosen by task, cost and sensitivity
  • Switch models without rewriting the integration
  • Models evaluated on your own cases

CLAUDE · GPT · MISTRAL · LLAMA

Questions from your IT team

Where is our data stored?

In the environment you choose. We do not host your data on our side.

Who controls what an agent does?

You do. Critical actions go through human approval, and all of them are logged.

Which AI models do you use?

Whichever fits your case and constraints. We compare several models on your test data.

How does the security review with our IT team work?

We provide the architecture, a map of data flows, and answer your security questionnaires.

Need a pack for your security committee?

We prepare the architecture and security documentation for your project with you.