Which of the following is a direct containment technique for AI incidents?

• A. Deploy data input/output validation.
• B. Ignore input validation to speed recovery.
• C. Increase system load to test resilience.
• D. Delay revoking access to datasets.

Answer: (A)

Direct containment in AI incident response means rapidly limiting impact by enforcing safeguards at the system boundary. Deploying data input/output validation directly reduces risk by filtering and validating what enters the AI system and what leaves it, preventing poisoned inputs, leakage, or unsafe outputs from propagating through the model. This action stops the problem at its entry/exit points, containing the incident at its source.

Ignoring input validation would allow harmful data to pass through, worsening the incident. Increasing system load is a resilience test or stress measure, not a containment action. Delaying revoking access to datasets prolongs exposure and undermines containment efforts.