A widespread view holds that agentic AI and AI agents fall through the regulatory gap and are not covered by the EU AI Act. The authors of the Unite.AI piece disagree: agentic AI is subject to the regulation because it falls under the broader definition of an AI system. They base this assessment on FAQs from the European Commission on "agents" and "agentic AI".
Background
According to the authors, the AI Act regulates not specific technologies but what a technology does and which risks it creates. That is why the regulation does not mention the word "agent". Agentic AI often builds on large language models, which are regulated as general-purpose AI models. If an agentic system interacts with people, the regulation's transparency requirements also apply.
If an AI model is classified as posing a systemic risk, the provider faces additional obligations, including assessments of the model's capabilities – including its autonomous and agentic functions. According to the piece, providers of systems that could affect fundamental rights must establish comprehensive risk management systems, which also assess how users could misuse the system.
Assessment
The authors warn of regulatory capture: if providers such as the companies named in the piece propose self-regulation, this could weaken enforcement of the regulation. As a comparison they cite the White House, whose superintelligence agreement is a "voluntary agreement". The piece itself acknowledges that the regulation is not perfect and will probably need further development.
For companies deploying AI agents, this assessment could mean that transparency and risk management obligations also apply to agentic systems, provided they fall within scope. Which obligations apply in a given case depends on what the system is used for; the piece does not replace that individual review.
Sources
Editorially owned by Ideal Syka. Sources and method: Newsroom & method. Tips and corrections: ai@i6eal.de.




