What Does the EU's AI Code of Practice Contain, and Why Won't Meta Sign It?

Imagine a world where artificial intelligence (AI) is growing faster than ever before, but Europe is trying to rein it in with rules. Something interesting is happening right now: Meta, the company behind Facebook and Instagram, has decided not to sign the voluntary EU AI Code of Practice. This move raises questions about the future of AI regulation in Europe. Let's take a detailed look at it, step by step, based on available information from relevant sources. Everything described here comes directly from verified details—no speculation, just facts.

What Happened Between Meta and the EU AI Code?

According to information from The Information, Meta has decided not to sign the EU AI Code of Practice. This document is a voluntary set of guidelines intended to help companies comply with Europe's new AI law, the AI Act. As a major player in the field of AI, Meta could have been one of its key signatories, but it chose the opposite approach. The headline of The Information article states clearly: "Meta Won’t Sign EU’s AI Code of Practice." This suggests that Meta sees potential problems in the Code that could hinder innovation. The headline itself and the context point to growing tensions between technology giants and European regulators.

This position is no coincidence. Joel Kaplan, Meta's Vice President, shared a LinkedIn post titled "Europe is heading down the wrong path on activity." In it, he warns that Europe could be taking the wrong approach to AI regulation, which could weaken the competitiveness of European companies compared with those in the US or China. Kaplan emphasizes that overly strict rules could cause Europe to fall behind in AI development. The post is part of a broader debate in which Meta expresses concerns about bureaucracy that could stifle innovation and create problems for developers.

What Exactly Is the EU AI Code of Practice?

Now we come to the heart of the matter—what is the much-discussed EU AI Code of Practice? This Code is a set of voluntary guidelines published by the European Commission on July 10, 2025. Its purpose is to help providers of general-purpose artificial intelligence models (GPAI—general-purpose AI) comply with the requirements of the AI Act, particularly during the transitional period from August 2025 to August 2027 or later, when binding standards will come into force.

The Code was developed through a multi-stakeholder process involving nearly 1,000 stakeholders—from independent experts and model providers to representatives of EU Member States and civil society. The process was supported by the European AI Office. It is not a binding law, but a voluntary tool that allows companies to demonstrate compliance with the AI Act. If a company signs it, this can reduce its administrative burden and provide greater legal certainty.

The Code is divided into three main sections:

• Transparency: Applies to all providers of GPAI models. It focuses on clearly communicating how models work, including the data used to train them.
• Copyright: Also applies to all providers and addresses how to respect copyright when training AI on data.
• Safety and security: This section applies only to models with potential systemic risk—that is, models that could threaten people's health, safety, or fundamental rights due to their broad applicability or high capacity.

The purpose of the Code is to create a "human-centric, trustworthy, and safe AI ecosystem" in Europe that supports innovation while protecting rights, democracy, and the rule of law. The European AI Office will assess the Code and may formally approve it. Companies such as Anthropic have already expressed support, but Meta is keeping its distance, underscoring the differences in approach.

Related Information and Broader Context

While researching Meta and the EU Code online, I found related details. For example, the Code is not definitive proof of compliance—companies must still meet all the requirements of the AI Act. It will be supplemented by further Commission guidelines on key GPAI concepts. Sources such as the European Commission's website (digital-strategy.ec.europa.eu) emphasize that the Code primarily helps with models posing systemic risks, where misuse could lead to major problems.

In the context of Kaplan's LinkedIn post, concerns have emerged that Europe could lose the global AI race. For example, while the US and China are investing heavily in AI without such strict regulations, the EU is trying to find a balance. As an American company, Meta sees this as a risk to its operations in Europe. Other sources, such as artificialintelligenceact.eu, confirm that the Code is crucial for the transitional period and supports transparency, safety, and respect for copyright.