AI-system rules still apply
Agentic architecture does not remove the need to classify the system under ordinary AI Act definitions and risk routes.
AIACT Research · Updated September 2026
AI agents regulation
AI agents are not a standalone legal risk category in the EU AI Act. Their obligations depend on the underlying system, intended purpose, capabilities, model and deployment context.
Tool use matters
Autonomous access to software and external systems can increase practical impact and should be documented in risk assessment.
Transparency
Agents interacting with people or generating content can trigger Article 50 duties.
High-risk context
An agent used for a sensitive Annex III use case may fall into high-risk obligations when the relevant provisions apply.
GPAI systemic risk
Autonomous capabilities can be relevant to systemic-risk assessment for general-purpose models.
Primary / official referenceOpen official source ↗
AIACT provides independent research and educational tools. This page is not legal advice. Regulatory status, guidance and implementation dates can change; verify the current official text before relying on it.