SECURITY BY DESIGN

AI understands the logic – not your secret credentials.

onflAI separates business configuration, AI validation, credentials and production execution. Secret credentials therefore remain outside the AI context.

01

Credential separation

Credentials are stored separately from the AI configuration and processed in encrypted form inside the platform for the technical connection.

02

Controlled approval

AI validates the configuration. Production activation takes place only after deliberate human review and approval.

03

Visible operations

Status, run history and errors remain transparent to responsible teams.

  • No credentials in the AI context
  • Validate before activating
  • Visible status and errors
04

Specific, not generic

The security architecture is a core product principle. Concrete regulatory requirements and deployment scenarios should be assessed for each business case.

05

MCP access with bounded permissions

Authentication, approved tools and permission scopes determine which actions an agent can invoke. Resources stay bound to their tenant and permission context. MCP access does not mean unrestricted access to enterprise systems.

06

Control over changes and data quality

Critical changes are reviewed and approved. Actions, status and errors remain traceable. Credentials are handled separately in the platform. ableX can be added for data quality and governance rules; the specific configuration is defined for your enterprise process.

07

What information AI needs

Configuration and validation use the business description and relevant structural information such as fields, mappings and rules. Secret credentials are not part of this context. Do not put passwords, API keys or tokens into free-text descriptions.

08

Credential separation is not a complete privacy assessment

Field names, sample data or a description can also contain confidential information. Review the content required and minimise it to business needs. Determine with the responsible teams which other data may be processed in your deployment.

09

What to review before approval

Are source, target, mandatory fields and write operations correct? Have missing values and failures been defined? Are permissions limited to what is necessary? Confirm expected results with test data. AI validation alone is not a security certification.

10

What belongs in your security review

Request the applicable information about hosting region, subprocessors, retention and deletion, key management, roles and agreed support for your deployment. This page does not replace contractual documentation or an assessment of your requirements.

  • Hosting & processing
  • Permissions & keys
  • Deletion & retention
  • Ownership & support
11

Clear boundaries instead of generic badges

These principles do not constitute a blanket certification, compliance or SLA commitment. The selected plan and confirmed technical and contractual terms for your deployment are authoritative.

Which business case should no longer become another IT project?

Register for free & start a business case →