FR · EN Book a demo
AI Desk Pro SaaS
Book a demo
Security and compliance

AI for your employees, built to clear your security review

AI Desk Pro runs on your Microsoft 365 tenant and on Microsoft Azure in Europe. Your data stays inside your governance perimeter, and your administrators keep visibility on every conversation.

Witivio is ISO/IEC 27001:2022 certified and publishes its commitments on a public Trust Center.

Book a technical session Open the Trust Center

30 minutes with a Witivio expert, no commitment.

Witivio is ISO/IEC 27001:2022 certified Witivio, Microsoft Partner Hosted on Azure Europe
In short

What is AI Desk Pro security built on?

AI Desk Pro is a SaaS solution hosted 100% on Microsoft Azure, in European regions, that signs every user in through Microsoft Entra ID and applies your tenant security policies. Witivio, the vendor behind it, is ISO/IEC 27001:2022 certified and publishes its commitments on a public Trust Center.

Security framework

How does AI Desk Pro protect your data?

Six control points, in the order your security team reviews them.

European hosting
The infrastructure is 100% Microsoft Azure, using managed PaaS services. Witivio operates no virtual machines. The regions used are North Europe or France Central. Traffic is encrypted in transit with enforced HTTPS and TLS, and data is encrypted at rest.
Identity and access
Employees and administrators sign in through Microsoft Entra ID. Your MFA, conditional access and organizational policies apply with no parallel setup. Consent is incremental: the application only requests the permissions it needs, profile read, read access to the SharePoint libraries you index, and directory access to add administrators.
Controlled knowledge scope
The agent answers only from the sources you give it, with an adjustable relevance threshold and a systematic citation of the documents used. Web search is off by default and limited to the sites you allow.
Content filtering and privacy
A content filter applies to inputs and outputs across four categories: hate and discrimination, sexual content, violence, and self-harm. Anonymous mode hides employee identity in the history your administrators review.
Oversight and traceability
Full conversation history with cited sources, filters by channel and by user, administrator review, feedback export, and an audit trail of knowledge base changes. Technical logs are retained for 90 days.
A certified organization
Witivio is ISO/IEC 27001:2022 certified. Security policy, subprocessors and architecture documentation are published on trust.witivio.com. A data processing agreement is available under GDPR.
AI models

Which models process your data, and under which terms?

ModelExecution providerTermsAvailability
OpenAI GPTAzure OpenAI Service (Microsoft)Microsoft product terms and Microsoft DPAAll plans, default model
Mistral LargeAzure AI Foundry (Microsoft)Microsoft product terms and Microsoft DPAPremium subscription
Claude OpusAnthropic API (direct call)Anthropic terms, activation requires explicit administrator consent. This is the only model not operated by Microsoft inside Azure.Premium subscription

The model is chosen agent by agent. You can stay exclusively on models operated by Microsoft inside Azure, with no outbound call to a third-party provider.

Architecture

How does AI Desk Pro fit into your Microsoft 365 tenant?

Your tenant stays the source of truth. AI Desk Pro only reaches your data within the permissions your administrators consent to.

Microsoft Azure managed services, deployed in the North Europe or France Central region.
  • Domains to allow: *.aidesk-pro.com, admin.aidesk-pro.com, docs.witivio.com, trust.witivio.com, and clarity.ms, used for usage measurement of the admin console. For the webchat: *.gpt-pro.com, *.botframework.com, and login.microsoftonline.com if you enable Microsoft 365 authentication.
  • Webchat: all traffic goes through a Bot Framework token that is only issued from the pages allowed in your configuration.
  • MCP server: delegated Entra ID authentication only. Every call runs on behalf of a signed-in user, with the same content filter and relevance threshold as in Microsoft Teams.
  • Accounts: tenant member users. Guest accounts are not supported. A service account is recommended for the SharePoint connection.
  • Synchronization: indexed documents are refreshed automatically. A document that has not changed is not re-indexed.
Diagram of the AI Desk Pro webchat authentication flow and Bot Framework token generation
Webchat authentication flow and Bot Framework token generation.
Governance

Your administrators stay in control

Three questions come up in every review: who administers, over what, and how you verify it.

Who
Named administrators, added from your tenant directory. Their rights are managed in the platform and can be revoked at any time from Microsoft Entra ID.
What
Sources, instructions, plugins, models and active channels are set agent by agent. Questions and answers go through a draft then a publication step, with an audit trail.
How
History, feedback, statistics and exports feed your periodic review. Conversations can be deleted on request.
AI Desk Pro conversation history filtered by channel, by agent and by period in the console
Filterable history: your administrators find a conversation, its sources and its rating.

Analytics, plugins and day-to-day administration are detailed on the AI Desk Pro features page.

Vendor review

What we provide for your security review

The documents below are available online or sent to your team on request.

ISO/IEC 27001:2022 certificate
The current certificate is shared through the Trust Center.
On request
Security policy and architecture
Security policy, organizational measures and architecture description.
Open the Trust Center
Data processing agreement
Data processing agreement (DPA) and subprocessor list, sent by our team.
On request
Product documentation
Authentication, limits, available models and content filtering, in open access.
Read the documentation
Security questionnaire
Our teams answer the vendor questionnaires from your procurement and security officers.
On request
Witivio commitments
The security framework that applies to all our products: hosting, encryption, GDPR and EU AI Act compliance.
Our security and trust commitments
Book a technical session Open the Trust Center
FAQ

Frequently asked questions about AI Desk Pro security

On Microsoft Azure, in European regions, North Europe or France Central, and exclusively on Microsoft managed services. Witivio operates no virtual machine for AI Desk Pro. Exchanges are encrypted in transit and data is encrypted at rest.

Indexing relies on the permissions you consent to, for example read access to the SharePoint libraries you designate. The scope is defined by your administrators and can be revoked at any time from Microsoft Entra ID.

Yes, in a history your administrators can review, with cited sources and the associated rating. Conversations can be deleted, and anonymous mode hides employee identity.

Yes. OpenAI GPT through Azure OpenAI Service is the default model, and Mistral Large also runs in Azure through Azure AI Foundry. Claude Opus calls the Anthropic API and is only enabled after explicit consent from your administrator.

Witivio applies GDPR in the design of the solution, offers a data processing agreement and is ISO/IEC 27001:2022 certified. Detailed commitments are published on trust.witivio.com.

Only member accounts of your tenant use the agent in Microsoft Teams, guest accounts are not supported. The webchat can stay open to the public or be restricted with Microsoft 365 authentication, depending on your configuration.

This page sums up the security framework in plain language: Azure hosting in Europe, Microsoft Entra ID authentication, encryption, ISO/IEC 27001:2022 certification, data processing agreement. The ISO certificate, the subprocessor list and the DPA are provided on request, and a Witivio expert can walk through these points with whoever owns the topic on your side, even without a dedicated security role.

Bring your security officer to the table

We run a dedicated technical session with your security team, with access to the Trust Center and the product documentation.

Book a technical session Open the Trust Center
AI Desk Pro mini-site

Go further

Continue your evaluation with the pages that matter most to your project.

Read the AI Desk Pro documentation (opens in a new tab)