NIST named agents as a standards object on 17 February 2026. CAISI's AI Agent Standards Initiative has three pillars: industry-led standards, open-source protocols, and security and identity research. It is a coordination point. It is not a finished protocol and not a procurement mandate.
The news
NIST posted "Announcing the AI Agent Standards Initiative for Interoperable and Secure Innovation" on 17 February 2026. CAISI, working with NIST's Information Technology Laboratory and interagency partners including the National Science Foundation, described three pillars:
- Facilitating industry-led development of agent standards and U.S. leadership in international standards bodies.
- Fostering community-led open source protocol development and maintenance for agents.
- Advancing research in AI agent security and identity to enable new use cases and promote trusted adoption across sectors.
The news release says NIST will announce research, guidelines, and further deliverables later, using convenings, RFIs, listening sessions, and other public-input tools. At launch CAISI pointed stakeholders to two comment paths: CAISI's Request for Information on AI Agent Security, due 9 March 2026, and ITL's AI Agent Identity and Authorization Concept Paper, due 2 April 2026.
CAISI said listening sessions on sector-specific barriers to AI adoption, with a focus on agents, would begin in April 2026. The Initiative page on nist.gov lists the program under CAISI's standards portfolio next to existing AI RMF materials.
The February announcement does not publish a finished agent protocol. It also doesn't publish a mandatory procurement standard. Vendors were already shipping MCP, A2A, and proprietary orchestration. NIST now has a federal desk for that work, with a security and identity track next to interoperability.
Who is bound
NIST CAISI and ITL authored the Initiative. Adoption is voluntary unless a contract or regulation cites NIST outputs. Agent-framework vendors, model providers shipping tool-use APIs, and enterprises piloting autonomous workflows are the audience, especially teams whose security reviewers already ask for NIST-aligned documentation.
Federal contractors putting agentic features into government-facing systems should track CAISI deliverables even without a mandate today. SaaS vendors selling inbox, ticket, or code agents into regulated accounts will see the Initiative name on 2026 questionnaires. That is orientation. It is not a pass/fail checklist yet.
What's new
Before February 2026, NIST's public AI work emphasized the AI Risk Management Framework and generative-AI profiles. CAISI now names agents as a distinct standards object, interoperability and security in one Initiative, and links the launch to an active RFI rather than waiting for post-incident guidance.
The mention of open-source protocol maintenance signals NIST will engage community specs, not only accredited-body standards. How NIST will reconcile overlapping industry protocols is UNKNOWN. The Initiative text does not pick winners.
What it does not settle
Dates: Initiative announced 17 February 2026; CAISI agent-security RFI comments due 9 March 2026; ITL identity concept paper due 2 April 2026; sector listening sessions from April 2026 per the announcement. Scope: voluntary federal guidance and standards facilitation, not an enforceable rule on private deployments. Cost: labor to answer RFIs and attend workshops; no NIST fee. Region: U.S.-centered coordination with stated intent to engage international standards bodies. Whether any agency will require Initiative-aligned protocols in 2026 procurements is UNKNOWN.
What to do now
Platform architects should map which agent protocols the product exposes and answer the RFI questions if they have field data on security failures. Enterprise procurement should add two questions: which external systems can the agent reach, and what identity model gates tool calls.