Europe softens AI rules after negotiating with big tech companies

EU countries and lawmakers reached a provisional agreement to significantly relax the artificial intelligence rules they originally proposed. The changes include postponing implementation of some rules and simplifying requirements that large tech companies had criticized as too strict.

The agreement, known as the third version of the AI Act, represents a victory for tech companies that lobbied intensively against the original regulations. The most significant modifications include:

• More permissive rules for general-purpose models: Companies developing models like Gemini, ChatGPT, or Claude will face fewer transparency and documentation obligations.
• Longer compliance deadlines: Some rules scheduled to take effect in 2026 have been postponed until 2028 to give companies time to adapt.
• Reduced infrastructure requirements: Auditing and reporting obligations have been simplified for models not considered "high-risk."

The criticism

Digital rights advocacy organizations and several academics noted that the agreement represents a significant concession to big tech power. "Europe started with the world's most ambitious AI law and ended up applauding the changes companies asked for," said a civil society organization involved in the negotiations.

Context

This agreement comes at a time when artificial intelligence has become a strategic priority for all major economic blocs. The U.S. is developing its own supervision framework (as the previous news item describes), and China already has its own rules. Europe, by softening its rules, sends a signal that it wants to be more attractive for AI investment.