Three US agencies—NIST, CISA, and the NSA—have jointly published a set of ten voluntary practices for handling incidents involving rogue AI agents. The guidance walks organizations through the entire incident lifecycle, from preparing for such events before they occur, to detecting and responding in the moment, and finally recovering and reporting afterward.
The practices are specifically tailored to the unique challenges posed by AI agents that act unexpectedly or maliciously, rather than traditional software incidents. While the document is voluntary, it signals a growing federal focus on AI security and gives organizations a common framework to align their own response plans.
Because the source is a single announcement, there are no differing viewpoints to compare. The emphasis is on proactive readiness and coordinated response, with the agencies offering the practices as a starting point rather than a regulatory mandate.