The European Commission formally launched its enforcement toolkit for the EU AI Act on August 2, ending the grace period that had shielded general‑purpose AI models from mandatory oversight. Under the new rules, the Commission can request detailed technical evaluations of any AI system before it is offered to European users, bar its market entry if it fails to meet safety standards, and levy penalties of up to €15 million or 3 % of a company’s worldwide annual revenue, whichever is higher.

Unlike earlier provisions that applied only to firms with an EU presence, the powers now reach any provider that makes a model available in the bloc, regardless of the company’s domicile. Non‑European developers must designate an authorised representative based in the EU, a move regulators describe as a “legal foothold” that prevents firms from evading scrutiny by operating from abroad.

Enforcement powers activated

Commission officials say the ability to demand model evaluations before regional release is the most significant new lever. “A US address does not put a lab outside the EU regulator’s reach,” noted Elisabetta Righini, a partner at Sidley Austin. She warned that liability extends beyond substantive breaches: “Refusing an information request, giving misleading answers, or blocking a model evaluation is finable on its own.” The Commission has already opened talks with OpenAI and Anthropic after recent cyber‑related incidents involving their models, according to Reuters.

Anthropic’s Mythos model, which the EU had pursued for months, finally became available for review in June, highlighting the agency’s growing leverage. The Commission can also issue “market access bans” that prevent a model from being sold or used in the EU until compliance gaps are remedied, a sanction that could disrupt revenue streams for companies that view Europe as a key market.

Industry response and geopolitical ripple

OpenAI’s European‑policy lead, Tom Gordon, said the company has “collaborated closely with the European Commission and the wider ecosystem on implementing the AI Act.” Google echoed that sentiment, pledging to remain “dedicated to meeting all applicable rules.” Both statements signal a willingness to engage, but the underlying tension is palpable.

The timing coincides with heightened U.S.‑EU friction over digital policy. In July, Brussels fined Google roughly $1 billion under the Digital Markets Act, prompting former President Donald Trump to threaten “substantial” tariffs on European goods. Adding AI‑specific enforcement to that mix gives Brussels another bargaining chip in negotiations with American tech giants.

Washington, for its part, is rolling out its own review mechanisms for frontier models, meaning U.S. labs now face parallel demands from two sides of the Atlantic. The dual scrutiny raises compliance costs and could force companies to redesign deployment strategies to satisfy divergent regulatory thresholds, timelines and penalties.

Legal experts caution that the new regime could set a global precedent. If the EU can effectively police AI models regardless of where they are built, other jurisdictions may follow suit, potentially ushering in a wave of extraterritorial tech regulations. For now, the Commission’s enforcement powers mark the most concrete step yet toward operationalizing the AI Act’s ambitions, and the tech industry is watching closely to see how the rules will be applied in practice.

This article was written with the assistance of AI.
News Factory APP - agentic news to boost your SEO & AEO.