Five types of agentic AI risk
Is this legally binding?
Guidance. Voluntary guidance. Best practice, not obligation, until a contract or a regulator cites it.
Because agents act in the real world, malfunctions cause real harm; the framework enumerates five risk types: erroneous actions, unauthorised actions, biased or unfair actions, data breaches, and disruption to connected systems.
From the source
“Erroneous actions; Unauthorised actions; Biased or unfair actions; Data breaches; Disruption to connected systems”
Section 1.2.2, Types of risk, p.7
What this connects to
2 relations. Official relations are the ones the source documents state; anything marked GAGE analysis is our reading, not an agency's.
Verified against the official source on 2026-08-17. GAGE is not affiliated with or endorsed by any agency named here, and nothing on this page is legal advice. How this is built and checked.