Skip to content
ajan.io

Control stays with you, at every layer.

You decide what agents can access, which actions they take, and where they wait for approval. Customer data is not used to train foundation models; each tenant's data stays in its own scope.

Six layers of control

Pick a layer

Who signs in and with which role.

  • Enterprise SSO and SCIM connection paths are documented and wired in when your environment is ready.
  • Users are managed with roles: who uses, who manages.

A document's journey

From upload to answer, a document passes a control at every step.

  1. Upload

    Encrypted in transit with TLS.

  2. Storage

    AES-256 class encryption at rest; tenant-owned scope.

  3. Search

    Only sources the tenant and agent can access are searched.

  4. Model call

    Data isn't used for model training; your own key on eligible models.

  5. Answer and record

    The answer shows its source; the action record is open to admins.

Who manages what?

Security is shared work. The platform provides the controls; your team decides how to use them.

  • Identity and access

    ajan provides
    Roles, SSO and SCIM connection paths
    You manage
    Who works with which role, your identity provider
  • Knowledge sources

    ajan provides
    Tenant-scoped storage and search
    You manage
    Which document is open to which agent
  • Actions

    ajan provides
    Automatic / approval / off setting
    You manage
    Each action's setting and its approver
  • Models

    ajan provides
    Model access policy, BYOK
    You manage
    Which model is open to whom, your own keys
  • Usage

    ajan provides
    Visibility and limits by user, agent, and model
    You manage
    Limit values and tracking
  • Retention and deletion

    ajan provides
    Processing deletion requests per the process
    You manage
    Retention period and legal process

Open status

We state plainly what's live, what's decided at setup, and what hasn't been published yet. No borrowed seals.

  • In the product
  • Set at onboarding
  • Not published
  • No model training on customer data

    Prompts, uploads, and outputs are not used to train foundation models.

    In the product
  • Encryption in transit and at rest

    TLS and AES-256 class encryption are baseline engineering controls.

    In the product
  • Approved actions and action records

    Approval at chosen steps; records open to admins.

    In the product
  • SSO and SCIM

    Connection paths are documented and wired to your identity provider at setup.

    Set at onboarding
  • Hosting region

    Decided with you at onboarding; this site doesn't invent a location.

    Set at onboarding
  • Independent audit certificates

    Shown here with the document once published.

    Not published
  • Service level commitment

    No published commitment; enterprise needs are discussed with sales.

    Not published

If you're running a security review

We can walk through these topics together as part of your pre-purchase security assessment.

  • Tenant isolation and access model
  • Identity provider connection (SSO, SCIM)
  • Action permissions and approval flows
  • Model providers and data flow
  • Retention, deletion, and records
  • Deployment options

Reporting a vulnerability

If you've found a vulnerability, choose the "Security and privacy" topic on the contact form. Don't paste passwords, keys, or personal data into the form; we'll get back to you for details.

Security questions

What security and privacy teams ask most.

All FAQs

No. Prompts, uploaded documents, and outputs are not used to train or fine-tune foundation models.

Start without giving up control.

Try a free workspace, or let's talk it through with your security team.