BRUSSELS -- The European Commission on Thursday released the final version of a voluntary code of practice for general-purpose artificial intelligence (GPAI), aimed at helping companies align with the European Union's AI Act.

The code offers guidance in three key areas: transparency, copyright, and safety and security. It applies to major GPAI models such as OpenAI's ChatGPT, Google's Gemini, Meta's Llama, and X's Grok.

Under the AI Act, GPAI models are defined as systems capable of performing a broad range of tasks and being integrated into various downstream applications. These models, typically trained on large and diverse datasets, support services like language generation and multimodal content creation.

The code was drafted by a panel of 13 independent experts and reflects input from over 1,000 stakeholders, including AI developers, academics, civil society organizations, rightsholders, and safety professionals, the Commission said.

The code will need to be endorsed by European Union (EU) member states before GPAI providers can voluntarily sign it. Signatories would benefit from reduced administrative burdens and greater legal certainty, according to the Commission.

However, a recent letter signed by over 50 business leaders, including representatives from ASML, Airbus, Mercedes-Benz, and French AI startup Mistral, urged the EU to delay the implementation of GPAI-related rules by two years. They warned that "unclear, overlapping, and increasingly complex" regulations risk undermining Europe's AI ambitions and could hamper the global competitiveness of European firms.

The EU's AI Act entered into force last year. Provisions related to GPAI, including governance and compliance requirements, will apply from Aug. 2.