NIST and CISA Publish Final Guidance on Protecting Identity Tokens

NIST and CISA have published a final interagency report offering new recommendations for securing identity tokens and assertions used by federal agencies and cloud service providers. The guidance responds to recent incidents involving token compromise and aims to strengthen protections for systems using single sign-on and API-based access.

The report, titled “Protecting Tokens and Assertions from Forgery, Theft, and Misuse,” was released on September 15, 2026. It provides implementation measures for organizations managing digital identity tokens, with a focus on limiting misuse and unauthorized access.

According to the guidance, access and identity tokens should have a maximum validity of one hour. The document states that expired tokens must be rejected by authorization services and policy enforcement points to prevent unauthorized access.

Recommendations also address key management practices. The guidance advises that signing keys for high-impact systems be rotated at least every 90 days, while keys for other systems should be rotated within a year. For systems with moderate impact or higher, keys are required to be stored in hardware-backed or isolated environments and not retained on servers, virtual machines, or containers.

What the numbers show

• Report released on September 15, 2026.
• Nearly 250 public comments incorporated into final version.
• Signing keys for high-impact systems to be rotated every 90 days.
• Tokens must not be valid for more than one hour.
• More than 60,000 emails were exfiltrated in a prior token compromise.

The guidance requires that tokens include an explicit audience field, and any access control system receiving a token without this field must reject it. Additionally, the report states that tokens and any personal data they contain should not be logged to avoid unintentional data exposure.

IR 8587 also notes that the recommendations apply to AI agents that use signed tokens to access systems, data, and APIs. However, the guidance does not address broader risks associated with AI agent access beyond the use of tokens.

The development of the report was informed by feedback from nearly 250 public comments and input from industry and interagency partners through the Joint Cyber Defense Collaborative. This included a technical exchange in June 2025 and a webinar in January 2026.

The report references two previous incidents as motivation for the new guidance: a 2020 supply chain intrusion involving forged SAML assertions and an incident in which foreign actors used a consumer signing key to forge tokens and extract over 60,000 emails from a federal agency. The document builds on updates to NIST SP 800-53 (Release 5.1.1) and Executive Order 14306, and offers architectural recommendations for identity providers and authorization servers, as well as improvements to key management and token verification processes.

* This article is based on publicly available information at the time of writing.