EU Draft Guidelines Define High-Risk AI Systems Under New Act

The European Commission has released draft guidelines aimed at clarifying which artificial intelligence (AI) systems are classified as high-risk under the EU AI Act. This classification is significant as it determines the extent of regulatory obligations that AI providers must adhere to, particularly concerning health, safety, and fundamental rights. Organizations can provide feedback on these guidelines until June 23, 2026.

The classification depends on two main criteria. First, AI systems that serve as safety components for regulated products, like automotive or aviation technologies, are inherently high-risk. Second, systems designed for specific use cases, such as biometrics, human resources, or emergency services, also fall into the high-risk category. This proactive approach makes sure that potentially harmful AI applications are closely monitored and regulated.

For those classified as high-risk, the EU AI Act imposes stringent obligations. Providers must implement broad risk management systems, enforce strict governance over training datasets, and establish post-market monitoring protocols. These requirements highlight the importance of clearly defining the intended purpose of AI systems, which is central to determining their risk classification.

Key Principles of High-Risk Classification

The draft guidelines span over 130 pages and include various examples illustrating which AI systems may be deemed high-risk. A critical takeaway is the focus on the intended purpose of the AI system. If the purpose is not explicitly limited to non-high-risk applications, it may be classified as high-risk. This means that vague disclaimers or generic exclusions in contracts will not suffice to avoid high-risk designation. Providers are encouraged to carefully craft promotional and contractual language to makes sure compliance.

Alongside the intended purpose, the guidelines specify that the combination of multiple AI systems into a single configuration will be assessed as one entity for classification purposes. This is particularly relevant for systems where outputs from various AI components influence decision-making processes, such as AI agents feeding data into a broader decision-making framework.

Implications for AI Providers

The publication of these draft guidelines indicates a broader intention by the European Commission to apply high-risk classifications comprehensively. As the AI field continues to evolve, compliance with these regulations will be key for providers aiming to operate within the EU market. The clarity provided through these guidelines is expected to ease regulatory navigation for AI developers and companies.

In light of these developments, companies should review their AI applications and makes sure that their compliance strategies align with the new guidelines. The proactive engagement period ending in 2026 also presents an opportunity for stakeholders to shape the final framework of the EU AI Act, potentially influencing how AI innovation can coexist with regulatory demands in the future.