← Back

Wire · regulatory

ChatGPT Didn't Break the AI Act. It Showed Why Adaptive Regulation Matters

Published

20 July 2026

Topic

regulatory

Sectors

AI Frontier Models

Geography

Europe

Source

Read at techpolicy.press

Verified

Fusion42 · 20 July 2026 · Fusion42 review

The EU AI Act adapted its original use-based regulatory framework to address general-purpose AI models after ChatGPT's emergence, introducing tiered obligations for GPAI providers. Enforcement begins August 2, 2026, with fines up to 3% of global annual turnover for non-compliance.

This Wire brief sits within Fusion42's coverage of AI Frontier Models, and 2 sources have reported it between 20 Jul 2026 and 26 Jul 2026.

◆ The Wire takeaway

If you ship a general-purpose AI model to Europe, the Commission now has real teeth: August 2, 2026 is when enforcement starts and fines up to 3% of global revenue become live. The rules won't change again; the Code of Practice is how you'll prove compliance, and models already live get until August 2027 to meet them.

Coverage

2 sources · first reported 20 Jul 2026 · latest 26 Jul 2026

Related on Wire

Topics

AI Frontier Modelseu-ai-actgpai-complianceenforcement-deadlineregulatory-adaptation