Record the incident and act to contain it: Following good post-deployment governance, what should the team do

The first step is to contain the harm and document the incident, its impact, and the response taken.

The question

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

1. Wait to see whether more users report the same behavior before deciding whether the event qualifies as an incident to log.

   Plausible as caution, but delaying containment of an active harm worsens the exposure.
2. Retrain the model on a larger dataset immediately, without first documenting the incident or containing the harm it is causing.

   Plausible as a fix, but retraining skips containment and documentation and takes far too long.
3. Delete the affected conversation logs to limit reputational exposure while the engineering team quietly investigates the underlying cause.

   Plausible as damage control, but destroying evidence undermines investigation and documentation duties.
4. Record the incident and act to contain it, capturing what happened, the impact, and the response steps taken to reduce further harm. ✓

   Correct: incident management calls for containing the harm and documenting the event, impact, and response as it unfolds.

How to remember it

The first step is to contain the harm and document the incident, its impact, and the response taken.

More Understanding How to Govern AI Development questions

• To proactively probe the system for weaknesses and failure: What is the primary goal of red teaming in this →
• Model or data drift: Which of the following is a recognized technical cause they should examine? →
• Technical documentation and instructions for use that tell: Which disclosure most directly serves that →
• All 426 Understanding How to Govern AI Development questions →