By Stuart Kerr, Technology Correspondent, LiveAIWire
Google’s EU AI Code of Practice decision became official on 30 July 2025, when Kent Walker, Alphabet’s president of global affairs and chief legal officer, confirmed in a company blog post that Google would sign the European Union’s voluntary compliance framework for general-purpose AI models. The move places Google alongside OpenAI, Microsoft, Anthropic, and Mistral AI as signatories, and it stands in direct contrast to Meta, which had refused to sign the same Code less than two weeks earlier, calling it regulatory overreach. Google’s signature came just three days before the EU AI Act’s general-purpose AI provisions took effect on 2 August 2025.
Walker’s post struck a notably qualified tone rather than a simple endorsement. He wrote that the final version of the Code came closer to supporting Europe’s innovation and economic goals than where it began, crediting the European Commission for incorporating industry feedback during the drafting process. At the same time, he stated Google remains concerned that the AI Act and the Code risk slowing Europe’s development and deployment of AI, specifically flagging departures from EU copyright law, approval delays, and requirements that could expose trade secrets as risks to European competitiveness.
What Signing Actually Commits Google To
The Code of Practice, finalised on 10 July 2025 by 13 independent experts working with the European Commission and informed by more than 1,000 stakeholders, is built around three chapters covering transparency, copyright, and safety and security. The transparency and copyright chapters apply to every general-purpose AI provider that signs, requiring published documentation of training data summaries and demonstrated policies for respecting copyright holders’ rights, including honouring requests not to use specific works in training data. The safety and security chapter applies only to the small number of providers, including Google, whose models are classified as carrying systemic risk under the Act.
Why Google Says the Economics Justify the Trade-Off
Walker’s post frames the decision partly in economic terms, stating that prompt and widespread AI deployment could boost Europe’s economy by 8 percent, or roughly 1.4 trillion euros annually, by 2034. That figure is Google’s own projection rather than an independent estimate, and it underpins the company’s argument that regulatory certainty, even with real compliance costs attached, is preferable to the market uncertainty of operating in Europe without a clear, government-endorsed path to compliance. Signing lets Google demonstrate AI Act compliance through the Code’s standardised documentation rather than through case-by-case negotiation with regulators, which the company frames as the more efficient route despite its stated reservations.
What This Means for How Gemini Operates in Europe
For users of Google’s AI products in the EU, the practical effect is increased documentation rather than any immediate change to how Gemini or other Google AI tools function day to day. Google must now publish clearer, standardised information about how its general-purpose models are trained, what data informed them, and how it handles copyright holder requests to exclude specific works from training data, information previously kept far more opaque. The transparency obligations are designed to give users, journalists, and regulators a clearer basis for understanding what a model like Gemini was built on, even though the underlying technology itself does not change as a direct result of the signing.
The Copyright Fight This Doesn’t Settle
Signing the Code does not resolve the deeper copyright tension underlying it. Our earlier reporting on the AI identity crisis explored how generative models blur the line between learned patterns and original expression, and the Code’s copyright chapter does not settle that argument so much as require companies to document their position on it. Publishers and rightsholders broadly welcomed the transparency requirement as a first step, but several creative industry groups have said the Code’s copyright provisions fall short of a binding legal remedy, since compliance is demonstrated through documentation rather than through independently verified licensing agreements.
Meta’s Refusal Makes This a Genuine Industry Split
Google’s decision is most notable in contrast to Meta’s. Where Google concluded that qualified cooperation served its interests better than open confrontation with EU regulators, Meta’s global affairs chief Joel Kaplan argued the Code introduces legal uncertainties that go beyond the AI Act’s actual legal scope. As detailed in our coverage of the EU AI Act’s earlier passage, the Commission has treated voluntary codes like this as a deliberate stepping stone toward enforceable regulation, following a similar path to how GDPR moved from guidance to binding law. Google’s signature strengthens that trajectory. Meta’s refusal is a visible marker of how much resistance remains among major AI developers to the EU’s specific approach, even when the underlying legal obligations apply to holdouts and signatories alike.
Whether Google’s more cooperative approach proves the better long-term strategy than Meta’s confrontation remains genuinely unresolved. Both companies must comply with the AI Act’s binding requirements regardless of whether they signed the voluntary Code, and full enforcement for systemic-risk models does not arrive until August 2027. What Google’s decision has already done is give the Commission a second major AI lab’s public buy-in, alongside its explicit acknowledgment that the arrangement is imperfect. That combination, qualified support rather than either full endorsement or outright rejection, may end up being the more common industry posture as more AI Act deadlines approach over the next two years.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, emerging technology, and their impact on business, society, and everyday life. LiveAIWire publishes original AI journalism every weekday at liveaiwire.com.