Key Points

1. On July 18, 2025, the European Commission released its General-Purpose AI Code of Practice and its Guidelines on the scope of the obligations for providers of general-purpose AI models under the AI Act 

   
   

2. Many companies have complained about the Code of Practice, and some have gone so far as to refuse to sign it—like Meta

   
   

3. Businesses who are in the European Union and who are outside but do business with the EU (see Article 2 regarding application) are recommended to review the AI Act, Code of Practice, and Guidelines and comply

Meta has just refused to sign the European Union’s General-Purpose AI Code of Practice for the AI Act. That’s right—Joel Kaplan, the Chief Global Affairs Officer of Meta, said in a LinkedIn post on July 18, 2025 that “Meta won’t be signing it”.

By general-purpose AI, I mean an AI model, including when trained with a large amount of data using self-supervision at scale, that displays significant generality and is capable of competently performing a wide range of distinct tasks regardless of the way the model is placed on the market and that can be integrated into a variety of downstream systems or applications. This does not cover AI models that are used before release on the market for research, development and prototyping activities.

What is the purpose of the AI Act?

As you may recall, section (1) of the Preamble of the AI Act states that the purpose is to:

“The purpose of this Regulation is to improve the functioning of the internal market by laying down a uniform legal framework in particular for the development, the placing on the market, the putting into service and the use of artificial intelligence systems (AI systems) in the Union, in accordance with Union values, to promote the uptake of human centric and trustworthy artificial intelligence (AI) while ensuring a high level of protection of health, safety, fundamental rights as enshrined in the Charter of Fundamental Rights of the European Union (the ‘Charter’), including democracy, the rule of law and environmental protection, to protect against the harmful effects of AI systems in the Union, and to support innovation. This Regulation ensures the free movement, cross-border, of AI-based goods and services, thus preventing Member States from imposing restrictions on the development, marketing and use of AI systems, unless explicitly authorized by this Regulation”

The AI Act classifies AI according to risk and prohibits unacceptable risk like social scoring systems and manipulative AI. High-risk AI is regulated, limited risk has lighter obligations, and minimal risk is unregulated.

The AI Act entered into force on August 1, 2024, but its prohibitions will be phased in over time. The first set of regulations, which take effect on February 2, 2025, ban certain unacceptable risk AI systems. After this, a wave of obligations over the next two to three years, with full compliance for high-risk AI systems expected by 2027 (August 2, 2025, February 2, 2026, and August 2, 2027 have certain requirements). Those involved in general-purpose AI may have to take additional steps (e.g., develop of Codes of Practice by 2025), and may be subject to specific provisions for general-purpose AI models and systems. See the timeline for particulars.

What is the Code of Practice for the AI Act?

The Code of Practice is a voluntary tool (not a binding law), prepared by independent experts in a multi-stakeholder process, designed to help industry comply with the AI Act’s obligations for providers of general-purpose AI models.

More specifically, the specific objectives of the Code of Practice are to:

• serve as a guiding document for demonstrating compliance with the obligations provided for in the AI Act, while recognising that adherence to the Code of Practice does not constitute conclusive evidence of compliance with these obligations under the AI Act

  
  

• ensure providers of general-purpose AI models comply with their obligations under the AI Act and enable the AI Office to assess compliance of providers of general-purpose AI models who choose to rely on the Code of Practice to demonstrate compliance with their obligations under the AI Act

  
  

Released on July 10, 2025, it has three parts:

1. Transparency: Commitments of Signatories include Documentation (there is a Model Documentation Form containing general information, model properties, methods of distribution and licenses, use, training process, information on the data used for training, testing, and validation, computational resources, and energy consumption during training and inference)

   
   

2. Copyright: Commitments of Signatories include putting in place a Copyright policy

   
   

3. Safety and Security: Commitments of Signatories include adopting a Safety and security framework; Systemic risk identification; Systemic risk analysis; Systemic risk acceptance determination; Safety mitigations; Security mitigations; Safety and security model reports; Systemic risk responsibility allocation; Serious incident reporting; Additional documentation and transparency

For each Commitment that Signatories sign onto, there is a corresponding Article of the AI Act to which it relates. In this way, Signatories can understand what parts of the AI Act are being triggered and complied with. For example, the Transparency chapter deals with obligations under Article 53(1)(a) and (b), 53(2), 53(7), and Annexes XI and XII of the AI Act. Similarly, the Copyright chapter deals with obligations under Article 53(1)(c) of the AI Act. And the Safety and Security chapter deals with obligations under Articles 53, 55, and 56 and Recitals 110, 114, and 115 AI Act.

In a nutshell, adhering to the Code of Practice that is assessed as adequate by the AI Office and the Board will offer a simple and transparent way to demonstrate compliance with the AI Act.

The plan is that the Code of Practice will be complemented by Commission guidelines on key concepts related to general-purpose AI models, also published in July. An explanation of these guidelines is set out below.

Why are tech companies not happy with the Code of Practice?

To start, we should examine the infamous LinkedIn post:

“Europe is heading down the wrong path on AI. We have carefully reviewed the European Commission’s Code of Practice for general-purpose AI (GPAI) models and Meta won’t be signing it. This Code introduces a number of legal uncertainties for model developers, as well as measures which go far beyond the scope of the AI Act.

Businesses and policymakers across Europe have spoken out against this regulation. Earlier this month, over 40 of Europe’s largest businesses signed a letter calling for the Commission to ‘Stop the Clock’ in its implementation. We share concerns raised by these businesses that this over-reach will throttle the development and deployment of frontier AI models in Europe, and stunt European companies looking to build businesses on top of them.

The post criticizes the European Union for going down the wrong path. It also talks about legal uncertainties, measures which go far beyond the scope of the AI Act, as well as stunting development of AI models and companies. There was also mention of other companies wanting to delay the need to comply.

To be sure, CEOs from more than 40 European companies including ASML, Philips, Siemens and Mistral, asked for a “two-year clock-stop” on the AI Act before key obligations enter into force this August.

In fact, the bottom part of the open letter to European Commission President Ursula von der Leyen called “Stop the Clock” asked for more simplified and practical AI regulation and spoke of a need to postpone the enforcement of the AI Act. Essentially, the companies want a pause on obligations on high-risk AI systems that are due to take effect as of August 2026, and to obligations for general-purpose AI models that are due to enter into force as of August 2025.

Contrastingly, the top of the document is entitled “EU Champions AI Initiative”, with logos of over 110 organizations that have over $3 billion in market cap and over 3.7 million jobs across Europe.

In response to the feedback, the European Commission is mulling giving companies who sign a Code of Practice on general-purpose AI a grace period before they need to comply with the European Union's AI Act. This is a switch from the July 10, 2025 announcement that the EU would be moving forward notwithstanding the complaints.

The final word appears to be that there is no stop the clock or pauses or grace periods, period.

New guidelines also released July 18, 2025

In addition, the European Commission published detailed Guidelines on the scope of the obligations for providers of general-purpose AI models under the AI Act (Regulation EU 2024/1689)—right before the AI Act’s key compliance date, August 2, 2025.

The goal is to help AI developers and downstream providers by providing clarification. For example, it explains which providers of general-purpose AI models are in and out of scope of the AI Act’s obligations.

In fact, the European Commission stated that “The aim is to provide legal certainty to actors across the AI value chain by clarifying when and how they are required to comply with these obligations”.

The Guidelines focus on four main areas:

1. General-purpose AI model

   
   

2. Providers of general-purpose AI models

   
   

3. Exemptions from certain obligations

   
   

4. Enforcement of obligations

The intention is to use clear definitions, a pragmatic approach, and exemptions for open-source. That said, the Guidelines consist of 36 pages of dense material that need to be reviewed and understood. For instance, the Guidelines answer the question, “When is a model a general-purpose AI model? Examples are provided for models in scope and out of scope.

What happens next?

As we can see from the above discussion, there are serious obligations that need to be complied with—soon. To that end, businesses in the European Union or who do business in the European Union (see Article 2 regarding application) are recommended to review the AI Act, the Code of Practice, and the Guidelines to ensure that they are ready for August 2, 2025.

After August 2, 2025, providers placing general-purpose AI models on the market must comply with their respective AI Act obligations. Providers of general-purpose AI models that will be classified as general-purpose AI models with systemic risk must notify the AI Office without delay. In the first year after entry into application of these obligations, the AI Office will work closely with providers, in particular those who adhere to the General-Purpose AI Code of Practice, to help them comply with the rules. From 2 August 2026, the Commission’s enforcement powers enter into application. And by August 2, 2027, providers of general-purpose AI models placed on the market before August 2, 2025 must comply.