Brussels set to push first EU AI Act checks on high‑risk uses and general‑purpose models this week

European Union regulators will open the week by moving from guidance to checks under the AI Act, with the European Commission’s AI Office and national authorities beginning coordinated requests for documentation from general purpose model makers and high risk deployers. The inquiries start Tuesday in Brussels, and target providers and users whose systems fall under Annex III categories like hiring and credit scoring, a person familiar with the Commission’s timetable said. The push matters because non compliance can trigger fines up to 35 million euros or 7 percent of global annual turnover, according to the regulation.

The first wave focuses on technical documentation, risk management, and systemic model transparency reports, which the Act requires for general purpose AI and for high risk applications now in force across the bloc. Enterprises that rely on cloud AI platforms will need model cards, training data summaries, and post deployment monitoring evidence ready for regulators, the Commission has said in prior guidance.

What changes this week

Providers behind widely used models are in scope. OpenAI’s GPT 4o released May 2024, Google’s Gemini 1.5 announced February 2024, Anthropic’s Claude 3 family launched March 2024, and Meta’s Llama 3 released April 2024, all count as general purpose AI under the law. The AI Office has signaled that models with significant compute and downstream reach face additional reporting, alongside voluntary codes that become benchmarks for enforcement.

Enterprises that embed AI in HR, lending, medical devices, and safety components must show conformity assessments and data governance. That will hit users of systems from SAP, Workday, Oracle, and Microsoft in EU operations. Workday reports more than 10,000 customers, and SAP says it serves over 100,000 enterprises, which underscores the breadth of affected deployments.

Cloud concentration raises the stakes. Synergy Research Group estimated in 2024 that AWS held roughly 31 percent of global cloud infrastructure, Microsoft Azure about 25 percent, and Google Cloud about 11 percent. Regulators are watching how foundation model access through these platforms shapes market power, the UK Competition and Markets Authority said in its 2024 updates on foundation models.

For technology leaders, this week is a line in the sand for AI governance. Procurement teams should lock vendor attestations, data residency, and incident reporting into contracts, and CISOs should align NIST AI RMF style controls with EU documentation demands. The compliance bar will influence model selection, cloud architecture, and budget planning for 2027, not just legal risk.