Back

NeoTeo Examines EU AI Act Rules for AI Agents

AI agents fall within the European Commission’s existing framework for AI systems and general-purpose AI models, rather than a distinct legal category, according to NeoTeo’s analysis published on November 20, 2026. The analysis describes the Commission’s considerations specific to agents as preliminary and examines transparency obligations alongside separate rules for high-risk systems and models carrying systemic risk.

NeoTeo uses the term AI agent for a system that takes information from its environment, processes it and acts on the result. Such systems often pair an interface with a general-purpose AI model, according to the analysis. It also notes that usage of the term varies and does not identify a single standard architecture.

The distinction between a complete system and its underlying model is central to NeoTeo’s account of the regulatory framework. Its analysis places high-risk classification at the system level, while identifying systemic-risk designation as a matter concerning the model.

For transparency, NeoTeo says Article 50 covers systems whose intended uses include interaction with people or content generation. Its account ties those requirements to the covered uses, not simply to whether a system is called an agent. The Commission’s non-binding guidelines for Article 50 give August 2, 2026, as the date from which the transparency provisions apply.

Under Article 6 of the AI Act, the high-risk classification concerns AI systems. The routes described in the regulation include use as a safety component in a regulated product requiring third-party conformity assessment, as well as inclusion in Annex III.

The regulation also sets conditions under which certain Annex III systems can qualify for an exception where they present no significant risk to health, safety or fundamental rights. Listing in that annex therefore does not invariably determine the classification. Under the Act, however, systems that profile natural persons retain high-risk status.

At the model level, the AI Act requires providers whose general-purpose AI models carry systemic risk to assess those risks and take mitigation measures. NeoTeo presents that obligation separately from the high-risk classification of an agent as an AI system.

* This article is based on publicly available information at the time of writing.