The Cybersecurity and Infrastructure Security Agency (CISA) and the National Institute of Standards and Technology (NIST) released Interagency Report (IR) 8587 on September 15. It hardens the identity tokens behind single sign-on (SSO), federation, and application programming interface (API) access. A stolen or forged token becomes a foothold.

Apply the same signed-token controls when artificial intelligence (AI) agents hold them. The report says agent access risks need further guidelines and, in places, new standards. National Cybersecurity Center of Excellence (NCCoE) agent-identity work is still early. This final is token hardening, not a full agent-authorization guide. (CISA; NIST; CSO)

Fortune on September 16 caught CIOs racing to put guardrails around agents. PwC's Joe Atkinson said "the agent made me do it" will not hold as a defense. Cisco's Thimaya Subaiya killed third-party assistants for one MyAgent platform. Workday's Gabe Monroy built an agent system of record for non-human identities.

ServiceNow sells an AI Control Tower. Zscaler's Sam Curry called agents a new form of insider. Inventories, short-lived credentials, and human approval for high-risk actions sit next to today's token checklist. (Fortune)

WSO2 made Agent Manager generally available September 17. Techweez covered it. Each agent gets a verifiable identity, Model Context Protocol (MCP) controls, and a sandboxed Kubernetes runtime. Gartner projects more than 150,000 agents per Fortune 500 company by 2028.

Only 13% of organizations say they have proper agent governance. Apache 2.0 self-host or managed cloud. Read it as product color next to IR 8587's identity gap, not a substitute for the federal token checklist. (Techweez)

Thursday already put Spain's Agencia Española de Protección de Datos (AEPD) first AI-agent personal-data breach notice on the desk. The organization says an agent logged in, hunted flaws, then changed records and opened invoices. The file is still under review.

You already have that morning. Today's file is a US token and identity final with an explicit agent-authorization gap. A Spanish General Data Protection Regulation (GDPR) notice is not IR 8587. (AEPD)

Wednesday already put Korea's Internet & Security Agency (KISA) AI Security Guide 2.0 on the desk. Seoul is still drafting privilege red lines for agentic and physical AI. That morning stays the Korea guide. The text is still open.

You already have that morning. Today's checklist is federal token lifecycle for SSO, federation, and APIs. A draft Korean privilege guide is not a CISA-NIST implementation report. (SEDaily)

Monday's Amodei essay is still the evaluator commit: permanent third-party reviewers with desks, badges, and employee-like access. Tuesday's Altman monitorability wall and multi-lab pact stay on the desk. Saturday's Cyber Resilience Act (CRA) Article 14 clock still runs: early warning in 24 hours, full notice in 72.

Those already had their mornings. They stay as the evaluator mechanism, the lab pact track, and the European product-incident duty next to today's token final. A signed-token guide does not replace desks or a 24-hour clock. (Amodei; TechTimes; European Commission)

Before the next identity or agent platform renewal, put four blanks on one page. Are tokens short-lived and revocable. Is there a separate inventory of non-human identities. Which high-risk agent actions need a named human. Are signed tokens kept out of logs and build pipelines.

IR 8587 is final for tokens and open on agent authorization, so write the questionnaire now. Put the same blanks next to every agent that can hold a token or call an API. (CISA)

The Move

CISA and NIST finalized a token-hardening guide for SSO, federation, and APIs. Same controls apply when agents hold signed tokens. The report itself leaves comprehensive agent authorization to further guidelines. Fortune's CIOs are still counting agents while that gap stays open.

Send legal, the CISO, and the identity owner a one-pager. Ask whether tokens expire and revoke, whether non-human identities sit in a separate inventory, which high-risk agent actions need a named human, and whether signed tokens stay out of logs and pipelines. Set those next to Thursday's Madrid blanks and Wednesday's Korea scopes.

If the answer is "we follow IR 8587, so agents are covered," the blank still goes next to every agent that can hold a token this week.

Seven bullets. One move.