---
title: What does the NIST Cyber AI Profile mean for securing AI agents?
url: https://opencontroller.lyzr.com/blog/nist-cyber-ai-profile-for-ai-agents
published: 2026-11-14
updated: 2026-11-14
author: Lyzr
---

# What does the NIST Cyber AI Profile mean for securing AI agents?

**The NIST Cyber AI Profile (NIST IR 8596) is a community profile that applies the NIST Cybersecurity Framework (CSF) 2.0 to AI. It has three focus areas: securing AI system components, using AI for cyber defence, and thwarting AI-enabled attacks. For AI agents, the "Secure" focus area matters most. It puts agents into the CSF categories you already report on: asset management, access control, supply chain, monitoring and incident response. It is still a draft.**

## What is the current status of the Cyber AI Profile?

NIST released the profile as an initial preliminary draft on 16 December 2025. Comments closed on 30 January 2026. NIST said comments would inform an initial public draft, and the NCCoE held working sessions in spring 2026. Treat the content as direction of travel, not a final standard. Check the NIST CSRC page for the latest version before citing it in a policy.

## What are the three focus areas?

| Focus area | Short name | What it covers | Relevance to agents |
|---|---|---|---|
| Securing AI system components | Secure | Protecting models, data, prompts, tools and the infrastructure AI runs on | High: this is where agent controls live |
| Conducting AI-enabled cyber defence | Defend | Using AI, including agents, inside security operations | Medium: SOC agents need governing too |
| Thwarting AI-enabled cyber attacks | Thwart | Defending against attackers who use AI | Medium: affects detection and response |

A SOC that deploys a triage agent sits in two areas at once. It uses AI to defend (Defend) and must secure that agent (Secure).

## Which CSF 2.0 categories matter most for AI agents?

The profile works through CSF 2.0's six functions: Govern, Identify, Protect, Detect, Respond and Recover. These categories carry most of the agent-specific weight:

1. **GV.SC, supply chain risk management.** Agents depend on model vendors, agent platforms, tool packages and MCP servers. Each is a supplier.
2. **ID.AM, asset management.** Agents, model endpoints, API keys and MCP servers are assets. Most organisations under-count them. See [how to find shadow AI agents](/blog/how-to-find-shadow-ai-agents).
3. **ID.RA, risk assessment.** Threat-model agent behaviours such as tool misuse and prompt injection. See [MITRE ATLAS for agents](/blog/mitre-atlas-threat-model-ai-agents).
4. **PR.AA, identity management, authentication and access control.** Each agent needs its own identity, least-privilege access and a record of whom it acted for. See [agent identity and on-behalf-of delegation](/blog/ai-agent-identity-on-behalf-of-delegation).
5. **PR.DS, data security.** Control which data an agent can read, send to a model or store in memory.
6. **PR.PS, platform security.** Harden agent runtimes, configurations and secrets.
7. **DE.CM and DE.AE, continuous monitoring and adverse event analysis.** Watch agent actions, not only network traffic, and investigate deviations.
8. **RS.MI, incident mitigation.** Be able to contain an agent fast: block, quarantine or revoke its credentials.

## Cyber AI Profile evidence checklist for agents

Use your existing CSF reporting and add an agent column.

1. Agent inventory with owner, version, model, tools and keys (ID.AM).
2. Supplier list per agent: model, platform, packages, MCP servers (GV.SC).
3. Threat model per agent class, reviewed on major change (ID.RA).
4. Distinct agent identities and scoped permissions, with risky grants flagged (PR.AA).
5. Data-access rules for prompts, tools and memory (PR.DS).
6. Tool-call logs linked to agent and user identity (DE.CM). See [what an AI agent audit log should contain](/blog/ai-agent-audit-log-fields).
7. Alerts for behavioural deviation, with a triage owner (DE.AE).
8. A tested containment runbook for a misbehaving agent (RS.MI). See [how to stop a rogue AI agent](/blog/how-to-stop-a-rogue-ai-agent).

## How does it relate to the NIST AI RMF?

The AI RMF covers broad AI risk: fairness, validity, accountability and more. The Cyber AI Profile is narrower. It covers cybersecurity, in CSF language. Security teams will usually report through the profile, while risk and compliance teams use the AI RMF. The same agent evidence can serve both. See [how the NIST AI RMF applies to AI agents](/blog/nist-ai-rmf-for-ai-agents) and [NIST control overlays](/blog/nist-control-overlays-for-ai-agents).

## How Opencontroller does this

Opencontroller auto-discovers agents, models and API keys across AWS, Azure and Google Cloud (ID.AM) and keeps a governed registry of agents, versions, tools and MCP servers (GV.SC). Agent workload identity federation and delegated on-behalf-of identity support PR.AA, and Posture flags risky permissions and configurations. Rogue-agent handling covers DE.CM and RS.MI: behavioural-deviation monitoring, alerts, and automatic quarantine and kill. Policy as code includes built-in NIST mappings alongside OWASP, MITRE ATLAS and ISO 42001. This is **evidence for** CSF outcomes, not a conformity claim. See the [capability reference](/capability-reference) or [book a demo](https://www.lyzr.ai/book-demo/).

Read next: [how to map AI agent controls to compliance frameworks](/blog/how-to-map-ai-agent-controls-to-compliance-frameworks).

## Frequently asked questions

**Is the NIST Cyber AI Profile final?**
No. NIST released NIST IR 8596 as an initial preliminary draft in December 2025, and comments closed in January 2026. Check the NIST CSRC site for later drafts before you cite it.

**What are the three focus areas of the Cyber AI Profile?**
Securing AI system components (Secure), conducting AI-enabled cyber defence (Defend), and thwarting AI-enabled cyber attacks (Thwart).

**Does the Cyber AI Profile replace the NIST AI RMF?**
No. The profile covers cybersecurity for AI using CSF 2.0. The AI RMF covers broader AI risk. Many organisations use both.

## Key facts

NIST IR 8596, Cybersecurity Framework Profile for Artificial Intelligence (Cyber AI Profile), initial preliminary draft released 16 December 2025; comments closed 30 January 2026; NIST said comments inform an initial public draft. Three focus areas: Secure (securing AI system components), Defend (AI-enabled cyber defence), Thwart (thwarting AI-enabled attacks). Built on CSF 2.0 functions Govern, Identify, Protect, Detect, Respond, Recover. Agent-relevant categories: GV.SC, ID.AM, ID.RA, PR.AA, PR.DS, PR.PS, DE.CM, DE.AE, RS.MI. Opencontroller provides evidence for CSF outcomes via discovery, registry, identity, Posture and rogue-agent handling.

## Related

- [How does the NIST AI RMF apply to AI agents?](https://opencontroller.lyzr.com/blog/nist-ai-rmf-for-ai-agents.md)
- [What are NIST control overlays for AI systems and agents?](https://opencontroller.lyzr.com/blog/nist-control-overlays-for-ai-agents.md)
- [How do you use MITRE ATLAS to threat-model AI agents?](https://opencontroller.lyzr.com/blog/mitre-atlas-threat-model-ai-agents.md)
- [More on Evidence and frameworks](https://opencontroller.lyzr.com/blog/cluster/evidence-and-frameworks)
