Security

Your data. Your control.

Deyxia agents work with sensitive data. Here is, in concrete terms, how it is protected and how you stay in control.

Encryption

Traffic between your browser, Deyxia and our AI model provider goes through encrypted connections (HTTPS/TLS).

  • HTTPS enforced across the entire platform
  • API keys kept on the server, never exposed to the browser
  • Passwords stored in hashed form

Access management

Accounts are created by your administrator or by our team. Nobody can sign up to your workspace on their own.

  • Named accounts, tied to your organization
  • Access removed in one click
  • Administration console restricted to administrators

Authentication

Sign in with email and password, or with your Google or Microsoft account when your organization enables it.

  • Limits on sign-in attempts
  • Error messages that don't reveal whether an account exists
  • Microsoft accounts linked by unique identifier, not just by email

Activity log

Sign-ins, missions, approvals and configuration changes are logged and can be reviewed.

  • Timestamp and author for every event
  • History of missions and their results
  • Export available on request

Permissions

Each role sees what concerns it: members use the agents, administrators manage accounts and plans.

  • Separate client and administrator roles
  • Missions and results visible only to their author
  • Console access limited to administrators

Human approval

An agent proposes, you decide. No outgoing action is carried out without your explicit approval.

  • An “Approve” button for every proposed action
  • Declines recorded in the log
  • Automations with a built-in approval step

Data isolation

Missions and results are tied to your account. Files given to an agent are not kept after the analysis.

  • Files processed in memory, then deleted
  • Only the mission result is saved
  • No mixing of data between organizations

Privacy

Your data is used only to run your missions. We do not sell it and do not use it for advertising.

  • Transfer to the model provider limited to the current mission
  • A reachable person in charge of the protection of personal information
  • Access and deletion requests handled on request

AI transparency

What the agent shows you, and what it doesn't do.

Every result comes with what you need to verify it. The model's internal reasoning is neither displayed nor stored: you see verifiable facts.

Steps taken

Reading, checks and calculations performed.

Sources used

Files, sheets, columns or sections consulted.

Actions taken

What was approved, declined, sent.

Result

Findings, key figures, limits of the analysis.

Frequently asked questions

Security and privacy.

Are my files used to train an AI model?

Deyxia does not train any model on your data. Files are sent to the model provider only to run the mission, and are not kept by Deyxia afterwards.

Who can see the results of my missions?

Only you, plus the platform administrators for support purposes. Other accounts have no access.

Can an agent act without my consent?

No. Outgoing actions (sending a report, recording a decision) wait for your explicit approval, and every decision is written to the log.

Do you hold a security certification?

We do not claim any certification at this time. We describe here the measures actually in place and are happy to answer your security questionnaires.

How do I exercise my rights over my personal information?

See our Law 25 page or write to us: we respond within 30 days.