The European Union has taken a major step toward regulating artificial intelligence by publishing its first draft of a Code of Practice for General-Purpose AI (GPAI) Models. This 36-page draft outlines guidelines for companies developing advanced AI systems, focusing on areas such as transparency, copyright compliance, and risk mitigation. Stakeholders are encouraged to provide feedback by November 28, with finalized rules expected by May 2025.
Regulating the Future of AI
General-purpose AI models, defined as those trained with computing power exceeding 10²⁵ FLOPs, are the focus of these new regulations. Companies like OpenAI, Google, Meta, Anthropic, and Mistral are among the likely targets of the guidelines, though the scope may broaden as AI technology evolves.
The draft emphasizes transparency in AI development, requiring companies to disclose information about the web crawlers used in training their models. This provision directly addresses concerns from copyright holders and content creators about unauthorized data usage.
Tackling Risks Head-On
Risk assessment plays a central role in the proposed code. The guidelines aim to prevent issues such as cyberattacks, discriminatory outputs, and potential AI “rogue” scenarios often dramatized in science fiction. To this end, companies must implement a Safety and Security Framework (SSF), which includes safeguarding model data, instituting robust access controls, and conducting continuous evaluations of risk mitigation strategies.
Governance for Accountability
The governance section of the draft pushes for corporate accountability. AI makers will be required to conduct regular risk assessments, involve external experts where necessary, and maintain transparency in their operations. This section underscores the EU’s commitment to ensuring that AI technologies are developed responsibly and ethically.
High Stakes for Non-Compliance
Companies that fail to meet the standards outlined in the AI Act face significant penalties. Fines can reach up to €35 million or seven percent of global annual profits, whichever is higher. These steep consequences aim to incentivize strict adherence to the guidelines.
For more information, read the full article on Engadget.