Document mitigations and post-deployment verification: Which documentation action best addresses the evidence
A risk register is incomplete when records do not connect identified risks with implemented mitigations and post-deployment verification.
The question
Assume EU AI Act high-risk duties apply to a media-production system that ranks generated headlines for editorial review. The team has a documented risk register and model-card description, but testing records do not show whether identified risks were mitigated after deployment. Which documentation action best addresses the evidence gap?
Preparing for AIGP? Take the free 5-min readiness quiz →
- Document mitigations and post-deployment verification against each identified risk. ✓Linking each risk to mitigation and verification fills the missing evidence while supporting risk management and post-market monitoring duties.
- Repeat the original pre-deployment accuracy benchmark without examining mitigations.A repeated accuracy benchmark may miss operational risks and cannot show whether documented mitigations were implemented or effective.
- Expand the model card with more architecture details.Architecture details improve technical description but do not demonstrate that identified risks were addressed through mitigation and follow-up testing.
- Have editors record their informal confidence in each headline.Editorial confidence may inform operations, but it does not evidence systematic mitigation or verification of documented high-risk controls.
The trap
Look for traceability: risk, mitigation, implementation evidence, and post-deployment verification should connect. How to remember it
A risk register is incomplete when records do not connect identified risks with implemented mitigations and post-deployment verification.
How many of these would you get right?
One of 1581 AIGP questions on Certsqill. Take a free five-minute check and see your score per domain — not one number, but which section to open tonight.
Test your AIGP readiness — freeMore Understanding How Laws, Standards and Frameworks Apply to AI questions
- Assess the retailer as a potential provider: What does this evidence most defensibly support? →
- Assess representativeness and label quality across: What evidence should it prioritize? →
- Name trained overseers and authorize pause: What additional control is most necessary? →
- All 394 Understanding How Laws, Standards and Frameworks Apply to AI questions →
Part of the Certsqill AIGP question bank · Understanding How Laws, Standards and Frameworks Apply to AI ·
Every answer, right and wrong, comes with its own explanation.
