Skip to main content

AI Shield

Core role

AI Shield is the control and protection layer around AI and partner interactions. It evaluates:
  • who is acting
  • what tool they want to use
  • which customer or arrangement is affected
  • whether consent exists
  • whether the action is allowed in the current jurisdiction
  • whether the action is allowed, challenged, approval-gated, or blocked

Typical AI Shield outcomes

  • allow
  • allow_with_redaction
  • allow_with_approval
  • challenge
  • block

Example controls

Data minimization

A customer service assistant can retrieve an account summary but not full document archives.

Action restriction

An AI assistant can prepare a payment but cannot release it without approval.

Jurisdiction gate

A tool can be available in one geography and blocked in another due to local rules.

Risk-based challenge

High-risk requests can require step-up authentication or human review.

Evidence model

AI Shield decisions are queryable by:
  • tool
  • actor
  • policy pack
  • outcome
  • affected business object
  • time window