AI Risk Wire

Controls

Undo for agent actions

Actions the agent takes are staged or can be reversed within a set window.

Layer L3 Application · Owner Deployer · Family Human oversight

Effective
Not effective
Not effective (bypassed)
Did not exist

By risk

RiskIncidents where the control was not effective or did not existIncidents where the control was effectiveIncidents where sources don't mention the control
Agent takes destructive or unapproved actions11
Unauthorised transactions or spend1
Cascading failure across agents and systems

Why it was not effective

No incident yet records this control as not effective.

Incidents where it was not effective or did not exist per month

0 so far in 2026; 0 in the same span of 2025.

Jan 2026: 0Feb 2026: 0Mar 2026: 0Apr 2026: 0May 2026: 0Jun 2026: 0Jul 2026: 0Aug 2026: 0Sep 2026: 0Oct 2026: 0Nov 2026: 0 Jan 26FebMarAprMayJunJulAugSepOctNovto date

19 of 183 incidents in Oct 2026 have been read for controls.

Added in reported fixes, and what followed

No reported fix names this control yet.

Frameworks

StandardClauseRequirement
NIST SP 800-53CP-10To verify when the text is stored
OWASP AgenticASI08To verify when the text is stored