The European Commission has revealed that the long-awaited Code of Practice to help companies comply with its landmark AI Act has been delayed until late 2025, marking a six-month setback from the original May 2 deadline. This guidance, aimed at developers of general-purpose AI models like ChatGPT, Google Gemini, and Mistral, is voluntary—but signing it grants legal certainty, while non-signatories risk scrutinized compliance.

Despite industry calls from tech giants like Google, Meta, and European firms including ASML and Mistral for a delay, the European Commission remains determined to maintain the Act’s timeline. Officials emphasize the importance of establishing harmonized, risk-based rules to ensure AI safety across the EU market.

The AI Act, which sets tough transparency, bias testing, and safety obligations, is scheduled to begin enforcement for general-purpose models on August 2, 2025, with full enforcement powers in 2026. However, the failure to release the compliance guide ahead of this deadline has prompted appeals for a “clock stop”—a potential pause—on the act’s implementation.

This delay reflects broader tensions between regulatory ambition and industrial readiness. Dozens of Europe’s largest companies have warned that rushed compliance could stifle innovation and deter investment. They argue that smaller businesses, in particular, could struggle to meet complex obligations without clear guidelines.

Meanwhile, Big Tech firms continue to push back. Reports suggest they exerted influence during the Code drafting process, prompting concerns that the guidelines may be diluted in their favor. Yet EU officials insist the final Code will be comprehensive—and committed to be released before enforcement begins.

As the clock ticks toward August, all eyes are on Brussels. Will the delayed guidance offer the clarity firms need, or will it lead to widespread calls for enforcement to be paused?