Define documented permitted-use boundaries enforced: Which control most directly addresses the risk?
Reducing secondary-use harm requires enforceable scope boundaries plus monitoring, not just disclosure, accuracy, or risk transfer.
The question
A company deploys a model validated only for screening applicants to one role, but other departments begin reusing it for promotion decisions and vendor selection. Leadership wants to reduce downstream harm from these secondary uses without banning the tool entirely. Which control most directly addresses the risk?
Preparing for AIGP? Take the free 5-min readiness quiz →
- Publish a transparency report describing the model's training data and intended purpose so internal and external stakeholders can understand its original scope.Plausible and useful for transparency, but disclosure alone does not prevent teams from continuing to reuse the model outside its validated context.
- Improve the model's predictive accuracy through additional fine-tuning so that it performs acceptably for each new department that has independently adopted the tool.Wrong because higher accuracy on unvalidated tasks legitimizes and expands the unintended use rather than controlling the downstream harm.
- Purchase broader liability insurance to transfer the financial exposure that the additional internal teams reusing the model create.Plausible as risk transfer, but insurance addresses financial fallout after harm occurs and does nothing to forecast or reduce the harm itself.
- Define documented permitted-use boundaries enforced by technical access controls that block deployment outside the validated context, and actively monitor for out-of-scope usage. ✓Correct because forecasting and reducing unintended-use harm requires binding scope limits, enforcement, and detection of drift into new uses.
The trap
Assuming documentation or transparency about a model's intended use is enough to prevent people from using it for something else. How to remember it
Reducing secondary-use harm requires enforceable scope boundaries plus monitoring, not just disclosure, accuracy, or risk transfer.
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 Deployment and Use questions
- Document each incident: Which documentation activity best fits this deployment obligation? →
- Pre-defined notification triggers: Which element is essential? →
- Region-scoped feature controls plus a documented: Which control best satisfies the ban without disrupting the →
- All 424 Understanding How to Govern AI Deployment and Use questions →
Part of the Certsqill AIGP question bank · Understanding How to Govern AI Deployment and Use ·
Every answer, right and wrong, comes with its own explanation.
