---
title: What does the EU Cyber Resilience Act mean for AI agent software?
url: https://opencontroller.lyzr.com/blog/eu-cyber-resilience-act-and-ai-agents
published: 2026-11-16
updated: 2026-11-16
author: Lyzr
---

# What does the EU Cyber Resilience Act mean for AI agent software?

**If you place agent software on the EU market as a product with digital elements, such as a downloadable agent, an SDK, a desktop assistant or a device with an embedded agent, the Cyber Resilience Act applies. Reporting of actively exploited vulnerabilities and severe incidents has applied since 11 September 2026. The full set of obligations, including essential cybersecurity requirements, applies from 11 December 2027.**

The Cyber Resilience Act (CRA) is Regulation (EU) 2024/2847. It entered into force on 10 December 2024. It regulates products, not organisations, and it applies to manufacturers wherever they are established.

## Does the CRA apply to AI agents at all?

It applies when the agent is part of a "product with digital elements": software or hardware made available on the EU market that connects directly or indirectly to a device or network.

| What you ship | Likely in CRA scope? | Why |
|---|---|---|
| Downloadable agent app or desktop assistant | Yes | Standalone software placed on the market |
| Agent SDK or framework sold or distributed commercially | Yes | Software product with digital elements |
| Device with an embedded agent | Yes | Hardware with embedded software |
| Local MCP server distributed to customers | Usually yes | Software that connects to other systems |
| Pure SaaS agent with no installed component | Generally no | SaaS is excluded unless it is remote data processing essential to a product's function |
| Internal agent built and used only in-house | No | Not made available on the market |
| Free open-source agent framework, no commercial activity | Lighter regime | Open-source software stewards have narrower duties |

If a SaaS agent backend is required for an installed product to work, the backend counts as remote data processing for that product. Many agent products have that shape.

## What must be reported since 11 September 2026?

Article 14 requires manufacturers to report two things: actively exploited vulnerabilities in their product, and severe incidents affecting its security. The timeline:

1. **Within 24 hours of awareness: early warning.**
2. **Within 72 hours: vulnerability or incident notification**, with more detail, unless already given.
3. **Final report:** for a vulnerability, within 14 days after a fix or mitigation is available; for a severe incident, within one month of the notification.

Reports go through ENISA's Single Reporting Platform to the CSIRT of the member state where you have your main establishment. These duties apply to products already on the market, not only new ones.

A severe incident is one that affects, or could affect, the product's ability to protect the availability, authenticity, integrity or confidentiality of data or functions, or that leads to malicious code in the product or in a user's systems. A prompt injection that makes your shipped agent run attacker instructions on a customer's machine fits that description.

## What applies from 11 December 2027?

The essential cybersecurity requirements in Annex I. For an agent product, the most demanding ones are:

- **Secure by default configuration.** Tool permissions, network access and data sharing start locked down.
- **Protection against unauthorised access.** Authentication for agent APIs and local servers.
- **Data minimisation.** The agent processes only data needed for its function.
- **Vulnerability handling.** A coordinated disclosure policy, security updates, and a software bill of materials covering at least top-level dependencies.
- **A support period** that reflects expected use, generally at least five years, with security updates throughout.
- **Technical documentation** and a conformity assessment before CE marking.

Agent products add dependencies most software does not have: model weights, prompt templates, tool definitions and MCP servers. Treat each as a component in the SBOM. See [what an AI bill of materials is](/blog/what-is-an-ai-bill-of-materials).

## How does the CRA interact with the EU AI Act?

Article 12 of the CRA says a high-risk AI system that meets the CRA's essential requirements is deemed to meet the AI Act's cybersecurity requirement in Article 15, as far as those requirements are covered. You still need the rest of the AI Act. Our guide to [EU AI Act obligations for AI agents](/blog/eu-ai-act-obligations-for-ai-agents) covers that side.

## What evidence should agent manufacturers build now?

1. **A component inventory** for each product version: models, tools, MCP servers, libraries.
2. **A vulnerability intake** that reaches someone who can file within 24 hours.
3. **Exploitation detection** in telemetry, so you know when a vulnerability is actively exploited.
4. **Red-team results** against prompt injection and tool misuse for each release.
5. **Default configuration records** showing what is locked down out of the box.
6. **A support-period statement** and update history.

## How Opencontroller does this

Opencontroller is for enterprises governing the agents they run, not a CRA conformity tool. It does produce evidence that helps teams shipping agents: a governed registry of agents, versions, tools and MCP servers; multi-turn and red-team evals; policy as code with mappings to OWASP, MITRE ATLAS, NIST and ISO 42001; and OpenTelemetry-based telemetry. Opencontroller Edge extends control to coding agents and local MCP servers on developer devices. See [what an AI agent control plane is](/blog/what-is-an-ai-agent-control-plane), or [book a demo](https://www.lyzr.ai/book-demo/).

## Frequently asked questions

**When did CRA reporting obligations start?**
On 11 September 2026. Most other obligations, including essential requirements, apply from 11 December 2027.

**Does the CRA cover SaaS AI agents?**
Pure SaaS is generally excluded. A cloud backend that is remote data processing essential to an installed product is in scope as part of that product.

**What are the CRA reporting deadlines?**
An early warning within 24 hours, a notification within 72 hours, and a final report within 14 days of a fix for vulnerabilities or one month for severe incidents.

## Key facts

The Cyber Resilience Act is Regulation (EU) 2024/2847, in force since 10 December 2024. Article 14 reporting of actively exploited vulnerabilities and severe incidents applies from 11 September 2026: early warning within 24 hours, notification within 72 hours, final report within 14 days of a fix (vulnerabilities) or one month (severe incidents), via ENISA's Single Reporting Platform to the relevant CSIRT. Most other obligations, including Annex I essential requirements and SBOMs, apply from 11 December 2027. It covers products with digital elements made available on the EU market, including standalone software; pure SaaS is generally excluded unless it is remote data processing for a product. CRA Article 12 links compliance to AI Act Article 15 cybersecurity for high-risk AI systems. Opencontroller provides registry, red-team evals and framework mappings as evidence. Opencontroller pricing starts from $250K per deployment per year.

## Related

- [What is an AI bill of materials (AI-BOM)?](https://opencontroller.lyzr.com/blog/what-is-an-ai-bill-of-materials.md)
- [What does the EU AI Act require for AI agents, and when?](https://opencontroller.lyzr.com/blog/eu-ai-act-obligations-for-ai-agents.md)
- [When does building an AI agent make you a provider under the EU AI Act?](https://opencontroller.lyzr.com/blog/when-does-building-an-ai-agent-make-you-an-ai-act-provider.md)
- [More on Evidence and frameworks](https://opencontroller.lyzr.com/blog/cluster/evidence-and-frameworks)
