EU AI Transparency Act Officially Takes Effect: All AI Systems Must Disclose Training Data Sources and Decision Logic

On July 1, 2029, the EU AI Transparency Act officially took effect. The legislation requires all AI systems operating in the EU market — from chatbots to autonomous vehicles to medical diagnostics — to disclose three core pieces of information: training data sources and composition, a summary of model architecture, and explainability reports for key decision logic.

The Act sets a six-month compliance window; all AI systems must submit transparency reports by January 1, 2030. Non-compliant companies face fines of up to 4% of global annual revenue.

EU Digital Policy Commissioner Henna Virkkunen said: "Citizens have the right to know how the AI affecting their lives works. Transparency is not optional — it's a fundamental right."

The Act immediately triggered tension across the tech industry. OpenAI, Google, and Meta have all indicated they will comply while calling for "tiered transparency" — different disclosure standards for different risk levels. OpenAI stated that fully disclosing GPT training data composition would involve trade secrets and copyright disputes.

Digital rights organization Electronic Frontier Foundation praised the legislation as an "AI regulation milestone" but expressed concern that the six-month window was too tight for small and medium enterprises.