Clearly document intended and prohibited uses: Which measure is most appropriate?

Documenting scope and adding guardrails and disclaimers steers users away from unsupported, high-stakes unintended uses.

The question

Preparing for AIGP? Take the free 5-min readiness quiz →

1. Clearly document intended and prohibited uses, and add usage guardrails and disclaimers that steer users away from unsupported, high-stakes applications like legal advice. ✓

   Correct because defining scope and adding guardrails directly reduces the unintended, high-stakes use.
2. Assume users will read the general terms of service and therefore take no additional steps to constrain the high-stakes legal use the translation tool is now being applied to.

   Wrong because relying on unread terms leaves the harmful, high-stakes unintended use unaddressed.
3. Expand the model's capabilities so it also handles legal questions competently, since a large share of users are already applying it for exactly that purpose.

   Plausible but wrong because building for the unsupported use raises the stakes instead of reducing the risk.
4. Restrict documentation of the model's limitations to avoid drawing attention to the use cases that the organization has not officially endorsed or validated.

   Wrong because hiding limitations deprives users of the guidance that would steer them away from misuse.

How to remember it

Documenting scope and adding guardrails and disclaimers steers users away from unsupported, high-stakes unintended uses.

More Understanding How to Govern AI Deployment and Use questions

• Maintain a structured incident log capturing the issue: Which documentation practice best meets this goal? →
• Timely, accurate, and consistent messaging to affected: Which feature is most important for that plan to be →
• A documented deactivation procedure with clear authority: Which control best enables this? →
• All 424 Understanding How to Govern AI Deployment and Use questions →