The EU AI Act: Implementation Status Report

Europe's landmark AI regulation is now in effect. The implementation is revealing where the rules work and where they do not.

Published: 28 July 2026 Category: AI Policy / Regulation Sources: [EU Commission reports, industry compliance data, July 2026]

The Framework

The EU AI Act entered full implementation in mid-2026, creating the world's most comprehensive AI regulatory framework. The Act classifies AI systems by risk level — minimal, limited, high, unacceptable — and imposes obligations proportionate to risk. High-risk systems require conformity assessments, risk management systems, and human oversight. Unacceptable systems — social scoring, manipulation, biometric identification in public spaces — are prohibited.

Six months in, the implementation picture is mixed. Some provisions are working as intended. Others are creating confusion, compliance costs, and competitive disadvantages that regulators did not anticipate.

What Is Working

Transparency requirements have been the clearest success. AI systems interacting with humans must disclose that they are AI. Deepfakes must be labelled. These rules are technically straightforward and broadly complied with. Consumers have more information about when they are interacting with AI, and while the disclosures are sometimes buried in terms of service, they exist.

The high-risk system database — a registry of AI systems used in critical domains like healthcare, transportation, and law enforcement — is operational and growing. Registration is mandatory, and the database provides visibility into where high-risk AI is deployed. This is genuinely useful for accountability, though the enforcement mechanisms for non-registration remain unclear.

What Is Not Working

The definition of "high-risk" is broader than industry expected. A system used for employee scheduling in a large company qualifies as high-risk if it "affects working conditions." Almost every enterprise AI system affects working conditions in some way. The result is that many companies are classifying systems as high-risk that regulators probably did not intend to capture, creating compliance burdens that favor large companies with legal departments over smaller competitors.

The conformity assessment process is slow and expensive. High-risk systems must be assessed by notified bodies — third-party auditors accredited by EU member states. The number of notified bodies is limited, the queues are long, and the costs are substantial. Some companies report spending six months and €200,000 on assessments for systems that pose minimal actual risk.

The Competitive Impact

The AI Act is affecting competition in predictable ways. American companies — with larger legal budgets and existing compliance infrastructure — are navigating the Act more easily than European startups. Chinese companies, particularly those offering open-weight models, are largely ignoring the Act, calculating that EU enforcement against foreign entities will be limited. European companies are caught in the middle: subject to the rules, without the resources to comply as easily as larger competitors.

The result is a potential irony: the AI Act, designed to protect Europeans from AI harms, may be protecting Europeans from European AI companies.

The Verdict

The AI Act is a genuine attempt to regulate a transformative technology before it transforms too much. That ambition is commendable. The execution is flawed but improvable. The European Commission has acknowledged implementation challenges and is issuing guidance to clarify ambiguous provisions.

Whether the Act becomes a global template or a cautionary tale depends on the next 18 months. If implementation can be streamlined — more notified bodies, clearer guidance, proportionate enforcement — the Act could demonstrate that AI regulation is possible without stifling innovation. If implementation remains bureaucratic and burdensome, the Act will serve as evidence that regulation and innovation are incompatible, strengthening the hands of those who oppose regulation everywhere.

The experiment is underway. The results are not yet in.

DEMYSTIFY