COSAiS is a NIST Computer Security Division project, created in July 2025, to write SP 800-53 control overlays for five AI use cases: adapting and using generative AI (assistants/LLMs), using and fine-tuning predictive AI, single-agent AI systems, multi-agent AI systems, and security controls for AI developers. A concept paper was released for comment on August 14, 2025, and an annotated outline for the predictive-AI overlay was posted on January 8, 2026 with initial feedback requested by February 13, 2026. No overlay had been finalized as of August 2026.
| Document | NIST COSAiS control overlays — Control Overlays for Securing AI Systems (COSAiS): SP 800-53 overlays for generative, predictive and agentic AI |
| Issued by | National Institute of Standards and Technology — AI Risk Management Framework |
| Type | Consultation |
| Status | Proposed |
| Published | Aug 14, 2025 |
| Applies to | Voluntary; organizations that use NIST SP 800-53 controls, including banks and their cloud and AI vendors subject to FedRAMP-style control baselines. |
| Official source | csrc.nist.gov ↗ |
| Use cases | Cybersecurity · Generative & agentic AI · Third-party & vendor AI · AI governance (general) |
What are the key points of NIST COSAiS control overlays?
- Overlays tailor the SP 800-53 Rev. 5 control catalogue to a specific technology, so organizations already assessed against 800-53 can extend existing control baselines to AI rather than adopting a new standard.
- Five planned overlays: generative AI assistant/LLM use; predictive AI use and fine-tuning; single-agent systems; multi-agent systems; and controls for AI developers.
- Concept paper released August 14, 2025; predictive-AI annotated outline released January 8, 2026 (feedback by February 13, 2026); the agent overlays are the least mature.
- NIST runs the project through a public Slack community and mailing list and expects to release discussion drafts use case by use case.
- Complements the AI RMF (governance), AI 100-2 (threat taxonomy) and IR 8596 (CSF profile) by supplying control-level detail.
What did NIST COSAiS control overlays change for banks?
COSAiS is the first NIST effort to specify AI security at the control level, which is what third-party risk and audit teams actually test. Once published, the overlays are likely to become the baseline banks require of AI vendors in contracts and the reference examiners use when asking how an LLM or agent deployment is secured; the single- and multi-agent overlays are the most relevant to the agentic-AI gap left by the 2026 model-risk guidance.
What is NIST COSAiS?
Control Overlays for Securing AI Systems: a NIST project, started July 2025, to produce SP 800-53 control overlays for generative AI, predictive AI, single-agent and multi-agent AI systems, and AI developers. It is still in draft.
Are the COSAiS overlays finished?
No. As of August 2026 NIST had released a concept paper (August 2025) and an annotated outline for the predictive-AI overlay (January 2026); the agent overlays remain under development.
| Date | Document | Status |
|---|---|---|
| Apr 7, 2026 | AI RMF critical-infrastructure profile (concept note) — Concept Note: AI RMF Profile on Trustworthy AI in Critical Infrastructure | Proposed |
| Jan 12, 2026 | CAISI RFI on AI agent security (2026) — Request for Information: Security Considerations for Artificial Intelligence Agents | Proposed |
| Dec 16, 2025 | NIST IR 8596 (Cyber AI Profile) — Cybersecurity Framework Profile for Artificial Intelligence (Cyber AI Profile), NIST IR 8596 — preliminary draft | Proposed |
| Mar 24, 2025 | NIST AI 100-2e2025 (Adversarial ML) — Adversarial Machine Learning: A Taxonomy and Terminology of Attacks and Mitigations (NIST AI 100-2e2025) | Final |
| Jul 26, 2024 | NIST AI 600-1 (Generative AI Profile) — Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile (NIST AI 600-1) | In force |
| Jan 26, 2023 | NIST AI RMF 1.0 — Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1 | In force |
Follow every document these regulators publish
6 curated AI stories for banking executives · Every morning · Free
Subscribe to BankingNewsAI →