Independent check of task results
Whether a task succeeded is checked against the system's actual state, not the agent's own report.
Layer L3 Application · Owner Deployer · Family Testing, governance and output integrity
By risk
| Risk | Incidents where the control was not effective or did not exist | Incidents where the control was effective | Incidents where sources don't mention the control |
|---|---|---|---|
| Deception, concealment or sabotage | 7 | ||
| Gaming the task (reward hacking) | 2 | ||
| Made-up output relied on | 1 | 6 | |
| Silent wrong state or drift |
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.
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
| Standard | Clause | Requirement |
|---|---|---|
| NIST AI RMF | MEASURE 2.5 | To verify when the text is stored |