EU AI Act General Purpose AI Enforcement Powers and Article 50 Transparency Obligations Go Live on August 2, 2026
The regulatory landscape for artificial intelligence in Europe officially entered its active enforcement phase on August 2, 2026. Under the landmark EU AI Act, the European Commission’s newly formed European AI Office gained direct, centralized supervisory authority and formal enforcement powers over providers of General Purpose AI (GPAI) models.
While GPAI providers have technically been obligated to comply with documentation, copyright policies, and training data disclosure requirements since August 2, 2025, the AI Office could not issue formal penalties or conduct mandatory audits during the initial transition year. That grace period has officially ended, exposing non-compliant firms to severe financial and operational sanctions.1
Enforcement Desk and Regulatory Tools
To support its new mandate, the AI Office has established an active Enforcement Desk staffed with approximately 40 new specialized hires. Starting in August 2026, the AI Office began deploying a robust suite of investigative tools, including:
- Requests for Information (RFIs) and Access (RFAs): Mandatory demands for technical documentation, risk assessments, and testing records.
- Model Evaluations: Direct, independent testing of frontier models to verify safety and capability claims.
- Physical Inspections: On-site audits of provider premises to verify compliance.
- Severe Financial Sanctions: Under Article 101 of the AI Act, the Commission can levy fines of up to €15 million (approximately $17.2 million) or 3% of global annual turnover (whichever is higher) for non-compliance with GPAI obligations.
Newly Opened Complaint Routes
On August 20, 2026, the EU AI Office dramatically accelerated its enforcement posture by opening three formal complaint routes. These channels allow downstream developers, integrators, and whistleblowers to report violations of Articles 53 to 55 (the core GPAI transparency and copyright compliance rules).
Early reports indicate that the AI Office is already utilizing these routes to evaluate complaint filings regarding Meta and Google systems.
Compliance Gaps and US Lab Exposure
The activation of these powers has created immediate legal friction for leading US artificial intelligence laboratories. While around 21 major providers—including Amazon, Anthropic, Google, IBM, Microsoft, Mistral AI, and OpenAI—have signed the GPAI Code of Practice (originally drafted in July 2025 and finalized in August 2025), signatory status does not exempt them from direct regulatory scrutiny.
A primary point of contention is training data transparency. Under Article 53, GPAI providers are legally required to publish detailed summaries of their training content. However, OpenAI's official EU AI Act compliance statement reportedly skipped these training data disclosures entirely. With the August 2 deadline now passed, this "copyright gap" is legally actionable, and the AI Office has indicated it will actively enforce these transparency mandates, signatures notwithstanding:
"What’s rarely appreciated is that GPAI liability isn’t limited to substantive breaches: refusing an information request, giving misleading answers, or blocking a model evaluation is fineable on its own." — Elisabetta Righini, partner at Sidley Austin, quoted in CNBC
US-based labs now face a critical compliance stretch. Providers who placed models on the EU market before August 2, 2025, have until August 2, 2027, to achieve full compliance, but any model introduced after August 2, 2025, is subject to immediate enforcement, creating a split regulatory environment for rapid-fire model releases.
-
An instance of Sovereign regulatory grace periods collapse the moment active enforcement bureaus levy their first statutory fines. — General Purpose AI developers face direct operational liabilities as the EU AI Office shifts from voluntary codes of practice to mandatory audits and active fine regimes. ↩︎