Data integrity becomes an architectural concern when autonomous systems can act without approval. An account-freezing agent may execute successfully while sending customers fabricated policy explanations or relying on compromised threat intelligence. section 1.1
The risk is not limited to inaccurate information. A poisoned input can cause an agent to freeze the entire customer base, turning a local detection into broad operational harm. section 1.1
Integrity questions for autonomous decisions include:
- Whether customer messages accurately reflect an existing policy. section 1.1
- Whether external information sources can be trusted or have been poisoned. section 1.1
- Whether authority boundaries and accountability limit the consequences of bad information. section 1.1




















