Define acceptance criteria for severe residual risks: What is the strongest conclusion?
Severe plausible harms and incomplete evidence require predefined acceptance criteria before release.
The question
A media company’s content-generation tool has low observed rates of ordinary errors, but a plausible failure could amplify false breaking-news claims. Testing is incomplete, the business wants immediate release, and no documented threshold defines acceptable residual information-integrity risk. What is the strongest conclusion?
Preparing for AIGP? Take the free 5-min readiness quiz →
- Delay all deployment until every possible failure is eliminated.Risk governance requires proportionate treatment and acceptance decisions, not an unsupported expectation that every possible failure disappear.
- Let the product sponsor accept risk using the current test summary.A sponsor may have a role, but current evidence and absent criteria cannot support a disciplined residual-risk acceptance decision.
- Define acceptance criteria for severe residual risks before release. ✓Incomplete testing and severe plausible consequences require explicit acceptance criteria before an accountable release decision can be made.
- Release because observed ordinary errors remain infrequent.Ordinary error frequency does not establish acceptability for a different, severe information-integrity failure mode.
The trap
A low general error rate does not answer whether a severe specific failure is acceptable. How to remember it
Severe plausible harms and incomplete evidence require predefined acceptance criteria before release.
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 to Govern AI Development questions
- Treat the severe scenario with proportionate controls: What response best reflects the low-frequency severe →
- Evaluate affected-person outcomes with efficiency before: What does the contrasting evidence most directly →
- Design an empowered human fallback: What does the combined evidence most directly support? →
- All 426 Understanding How to Govern AI Development questions →
Part of the Certsqill AIGP question bank · Understanding How to Govern AI Development ·
Every answer, right and wrong, comes with its own explanation.