A document should not give an assistant new permissions

An assistant may need to read a brochure, but the brochure should not expand its authority. NIST’s prompt-injection discussion prompts practical questions about source material, limited access, and reviewable actions.

Availability
Published
For
Readers

What happens next

The March 2025 NIST report on adversarial machine learning explains how outside material can try to redirect an AI system. The report is a technical taxonomy and discussion of mitigations, not a security assessment of genus. Its relevance for a service team is understandable: material an assistant reads should not acquire the authority to change what the assistant is allowed to do.

Source

Read the source at genusINC ·