NIST, NSA and CISA Issue Super Intelligence Incident Guidance

Oct 19, 2026
Three US agencies published ten voluntary practices for managing incidents involving advanced AI agents during training and evaluation.
NIST, NSA and CISA Issue Super Intelligence Incident Guidance

National Institute of Standards and Technology, the National Security Agency and the Cybersecurity and Infrastructure Security Agency (CISA) published ten voluntary practices for managing incidents during super intelligence training and evaluation, NIST announced on its website. The guidance covers preparation, detection, response, recovery and reporting.

The agencies define a control incident as significant uninstructed action by one or more agents that compromises policies, controls or security measures intended to limit a task or maintain human oversight. The document primarily applies to developers and evaluators running capable agents on extended tasks with limited supervision.

According to a LegalTech Daily analysis of the 14 page document, the practices call for organizations to retain evidence needed for investigations, including model weights, checkpoints, rejected tool calls, timestamps and records from sandbox and network activity. They also recommend tamper resistant logs, credential revocation, rebuilding compromised environments from verified baselines and testing containment systems before restarting agents. The guidance does not specify how long organizations should retain incident records.

We hope you enjoyed this article

Free newsletter

AI Policy Brief

Weekly report on AI regulations, safety standards, government policies, and compliance requirements worldwide.

Whitepaper

Tensordyne Napier: What If One Rack Could Do the Work of Nine?

Tensordyne

This Tensordyne whitepaper presents Napier, an inference-focused AI processor and rack-scale system based on the company’s TDN Math logarithmic number system. It examines infrastructure requirements for large mixture-of-experts and agentic models, compares major inference architecture approaches, and details the TDN AIP processor, TDN72 pod, TDN Link fabric, and Napier Ultra configuration. The paper reports simulation-based performance, cost, and accuracy-validation results, including Tensordyne’s projected comparison of one Napier rack with a nine-rack Nvidia Rubin plus Groq deployment; the chip is reported as taped out and in fabrication.

Read more
Free, six days a week

Daily AI Brief: the AI news that matters, in your inbox.