Ok, alright, fine: Google signs EU’s AI code of practice after initial opposition

Google has confirmed it will sign the European Union’s AI Code of Practice, a voluntary framework it previously opposed as too restrictive. The decision positions Google to influence AI regulation implementation while competitors like Meta refuse to participate, potentially giving Google strategic advantages in navigating Europe’s evolving AI legal landscape.

The big picture: Google’s reversal reflects a calculated shift from resistance to engagement, as the company seeks to shape rather than simply comply with EU AI regulations.

• The tech giant initially opposed the code for being too harsh but now believes its input has helped create a framework that could provide Europe with “secure, first-rate AI tools.”
• Google claims widespread adoption of AI tools could boost the EU economy by 8 percent (about 1.8 trillion euros) annually by 2034.

What Google must do: Under the code’s terms, the company will face new disclosure requirements and compliance measures.

• Google must publish summaries of its model training data and disclose additional model features to regulators.
• The code includes guidance on managing safety and security in compliance with the AI Act, which came into force last year.
• The framework provides paths to align model development with EU copyright law, a contentious issue for AI companies.

Who’s in and who’s out: Major tech companies are taking divergent approaches to the voluntary agreement.

• Meta has steadfastly refused to sign, claiming the code could impose too many limits on frontier model development as it pursues its “superintelligence” project.
• Microsoft is still considering whether to sign the agreement.
• OpenAI has signaled it will sign the code.

What they’re saying: Kent Walker, Google’s head of global affairs, expressed cautious optimism while maintaining concerns about potential innovation barriers.

• Walker noted that the code could stifle innovation if not applied carefully, something Google hopes to prevent through its participation.
• He says Google remains concerned that tightening copyright guidelines and forced disclosure of possible trade secrets could slow innovation.

Why this matters: The stakes extend far beyond voluntary compliance, as all AI companies operating in Europe must abide by the comprehensive AI Act regardless of code participation.

• The AI Act includes the world’s most detailed regulatory framework for generative AI systems, banning high-risk uses like intentional user manipulation and real-time biometric scanning in public spaces.
• Companies violating the AI Act face fines up to 35 million euros ($40.1 million) or 7 percent of global revenue.
• Companies adopting the voluntary code will enjoy lower bureaucratic burden and easier compliance with the AI Act.

The broader context: Europe’s proactive regulatory approach contrasts sharply with the U.S., where the current administration is actively working to remove existing AI limits and even attempted to ban state-level AI regulation for ten years in a recent tax bill.