The National Institute of Standards and Technology (NIST) plans to update its core incident-response guidance and establish a stakeholder community to help organizations respond to cyberattacks, malfunctions, and harmful actions involving artificial intelligence (AI) systems.
The road map calls for adapting NIST Special Publication (SP) 800-61, developing guidance for incidents involving autonomous agents, and examining how monitoring, logging, and information-sharing practices need to change. The work would help organizations address both attacks targeting AI systems and harm caused by their misuse or unintended behavior.
The planned efforts build on a workshop convened in response to the federal AI Action Plan’s directive for NIST and its Center for Advancing Innovation and Standards for Super Intelligence to collaborate with AI and cybersecurity stakeholders on standards, response strategies, best practices, and technical tools for AI incident management.
The Trump administration refers to AI as super intelligence (SI).
NIST Internal Report 8621, which summarizes the workshop’s findings and next steps, describes the workshop as the start of an ongoing process to assess existing practices, identify gaps, and inform future guidance. More than 400 participants joined from government, industry, academia, standards organizations, and civil society.
Updating response guidance
NIST plans to update SP 800-61 to address AI-specific considerations. The agency is considering a Part A focused on cybersecurity incidents involving AI systems and a possible Part B addressing misuse, malfunction, or other unintended actions that may cause harm.
Related guidance would cover incidents involving training and testing data, models, autonomous agents, supporting infrastructure, and the AI supply chain. A separate document would offer voluntary principles for managing incidents in which highly capable agents take unintended actions outside an operator’s intended task scope during training and evaluation.
The distinction reflects a broader response challenge: A technically secure system can still cause harm through incorrect outputs, unsafe decisions, or unintended actions. Workshop participants also warned that automated attacks and agent failures may unfold faster than human responders can react.
“Consequently, traditional incident response life cycles built for conventional software systems may fracture or be incomplete when applied to SI, making the updating of incident response in the presence of SI a priority,” the report states.
Building a stakeholder community
NIST plans to establish an SI Incident Management Community of Interest to identify emerging challenges, exchange lessons learned, and inform future publications. Workshops, tabletop exercises, and community discussions could help the group identify priority gaps and shape targeted guidance.
The agency also plans to convene Task Group 8, focused on agent capabilities, monitoring, and containment, through the NIST AI Consortium. Organizations developing and evaluating highly capable agents on long-horizon tasks would discuss incident-response challenges and strategies.
Strengthening monitoring and evidence sharing
Other work would examine how monitoring and logging practices should evolve to support AI incident detection, investigation, and response. That effort could inform updates or additional resources for SP 800-92, NIST’s log-management guidance, covering model life cycle events, telemetry, provenance, and forensic evidence.
NIST also plans to examine AI-specific information-sharing and reporting needs, including how existing guidance could address technical evidence and incident context. Separate work could establish common concepts and operational practices for tracing and observing AI system behavior.
Workshop participants stressed that responders need visibility into deployed systems, clear responsibilities across vendors and other partners, and access to useful evidence. They also identified a tension between preserving prompts, outputs, and tool-use records for investigations and protecting sensitive information those records may contain.
NIST said the scope and specifics of its guidance will evolve with stakeholder feedback and emerging needs. The road map does not set deadlines for the proposed deliverables.