AI Act: applicability of rules on general-purpose AI models

The applicability of the AI Act is going ahead as planned. After some hints and debate prior to the summer on whether or not there would be an extension of deadlines, the European Commission has confirmed that there will not be any delay or stop-the-clock. 

As a result, on 2 August 2025, obligations relating to general-purpose AI models and general-purpose AI models with systemic risk have become applicable. The obligations apply to models that are placed on the market after this date. For models that are already on the market prior to 2 August 2025, there is a grace period for providers until 2 August 2027 to ensure the general-purpose AI models (with systemic risk) are in compliance with the AI Act. As of 2 August 2026, enforcement actions are also possible.

Meanwhile, on 18 July 2025 the European Commission has issued guidelines on the scope of obligations for general-purpose AI models. It provides clarification on what a general-purpose AI model is and what the difference is with a general-purpose AI model with systemic risk, what it means to place on the market a general-purpose AI model (with systemic risk), who the provider is (in particular in case of modifications by downstream actors), what the exemptions relating to open source models entail and how the European Commission looks at the enforcement of the rules and obligations. 

In addition, the European Commission has adopted a Code of Practice, the adherence to which can assist providers of general-purpose AI models (with systemic risk) in demonstrating compliance with the AI Act, while harmonized standards are being developed. The Code of Practice consists of three chapters: transparency, copyright, and safety and security. Adherence to the Code of Practice is voluntary. Well-known companies, like Microsoft, IBM, OpenAI, and Amazon, are signatories to the Code. Notably missing is Meta.