Which control BEST mitigates the risk that an autonomous AI agent will take irreversible actions with significant consequences before a human can intervene?
Select an answer to reveal the explanation.
Short Explanation and Infographic
Here's the deal — b is correct because the minimal footprint principle — a key control from agentic AI safety frameworks — restricts agents to requesting only necessary permissions, preferring reversible over irreversible actions, and pausing for human confirmation before taking high-stakes or irreversible steps. A provides monitoring visibility but does not prevent harm.
Full explanation below image
Full Explanation
B is correct because the minimal footprint principle — a key control from agentic AI safety frameworks — restricts agents to requesting only necessary permissions, preferring reversible over irreversible actions, and pausing for human confirmation before taking high-stakes or irreversible steps. A provides monitoring visibility but does not prevent harm. C improves accuracy but does not add a control gate. D addresses security isolation, not action reversibility.