Google has confirmed that it will sign the European Union's general purpose AI code of practice, a voluntary framework intended to help AI developers prepare for compliance with the bloc's AI Act. The decision puts Google on a different path from Meta, which earlier this month said it would not sign the code.
Google signs on while keeping reservations
The code of practice is designed as a practical bridge between AI developers and the wider obligations created by the AI Act. By signing, companies agree to follow guidelines around documentation, training data and requests from content owners.
Google's commitment was described by Kent Walker, president of global affairs at Google, in a blog post on Wednesday. Walker said the final version of the code was better than the EU's initial proposal, but he also made clear that Google still has concerns about both the AI Act and the code.
We remain concerned that the AI Act and Code risk slowing Europe’s development and deployment of AI. In particular, departures from EU copyright law, steps that slow approvals, or requirements that expose trade secrets could chill European model development and deployment, har ming Europe’s competit iveness,
That position captures the balance Google is trying to strike. The company is choosing to participate in the EU framework, while also warning that implementation details could affect AI development, deployment and competitiveness in Europe.
What the EU AI code asks companies to do
The general purpose AI code of practice is voluntary, but it is tied to a larger regulatory system that is becoming more important for major AI companies. Signing the code means agreeing to a slate of guidelines that are meant to help AI providers align their processes and systems with the AI Act.
According to the source article, the guidelines include several clear expectations:
- Providing updated documentation about AI tools and services.
- Not training AI on pirated content.
- Complying with requests from content owners who do not want their works used in datasets.
These obligations matter because general-purpose AI models can be used across many products and services. A model may not be built for one single narrow application, so regulators are focusing on how providers document, manage and train these systems before they are deployed broadly.
The code also shows how copyright, transparency and trade-secret concerns are becoming central to AI regulation. Google is accepting the framework, but its comments show that the company sees risk in rules that it believes could slow approvals or expose sensitive business information.
Why August 2 is the key date
Google's announcement comes days before rules for providers of “general-purpose AI models with systemic risk” go into effect on August 2. Companies likely to be affected include Anthropic, Google, Meta, and OpenAI, along with several other large generative models.
Those companies will have two years to comply fully with the AI Act. That creates a period in which AI providers must prepare systems, documentation and risk-management processes while the EU framework continues to shape how compliance will work in practice.
The timing also explains why the code of practice has become a visible test of industry cooperation. Signing the code does not remove every concern companies have about the AI Act, but it does signal a willingness to work within the EU's compliance structure.
Meta has taken a different position. Earlier this month, it said it would not sign the code, calling the EU's implementation of its AI legislation “overreach” and saying that Europe was “heading down the wrong path on AI.”
The wider AI Act framework
The EU's AI Act is described as a risk-based regulation for AI applications. It bans some “unacceptable risk” use cases, including cognitive behavioral manipulation and social scoring.
The act also defines “high-risk” uses. The source article identifies examples such as biometrics and facial recognition, along with the use of AI in domains like education and employment.
For developers, the act is not only about banning specific use cases. It also requires them to register AI systems and meet risk- and quality-management obligations. That means compliance is likely to involve both product decisions and internal governance processes.
For major AI companies, the practical question is how to keep building and deploying systems while meeting these rules. Google's decision to sign the code suggests that it will engage with the EU process, even as it continues to argue that some requirements may create friction for European AI development.
What Google's decision signals
Google's move does not settle the debate over the EU AI code of practice. It highlights it. One major company is signing while raising concerns; another major company has refused to sign and criticized the EU's approach.
That split points to the central tension around AI regulation in Europe. Regulators are trying to set rules for powerful general-purpose AI models, while companies are weighing compliance, copyright, approval processes, trade secrets and competitiveness.
For now, the most concrete fact is that Google will sign the code before the August 2 rules for providers of general-purpose AI models with systemic risk take effect. The next phase will be about how companies turn that commitment into processes that satisfy the AI Act over the two-year compliance window.