European Union artificial intelligence (AI) rules, hailed as a world first, are facing a critical moment as negotiators attempt to finalize the details this week. The EU's AI Act, first proposed in 2019, was expected to be the first comprehensive AI regulation in the world, reinforcing the bloc's position as a global trendsetter in regulating the tech industry. However, negotiations have been complicated by the emergence of generative AI, which produces human-like work. Big tech companies are lobbying against what they perceive as overregulation, while European lawmakers are pushing for added safeguards for cutting-edge AI systems developed by these companies.

The race to establish guidelines for AI technology has intensified globally, with the US, UK, China, and coalitions like the Group of 7 major democracies all joining in. Researchers and rights groups have warned of the existential threats posed by generative AI to humanity and everyday life. Nick Reiners, a tech policy analyst at the political risk advisory firm Eurasia Group, expressed concerns that the AI Act may not be agreed upon before the European Parliament elections next year due to the complexity of the negotiations.

When the European Commission unveiled the draft of the AI Act in 2021, it focused mainly on product safety legislation and did not extensively cover general purpose AI systems like chatbots. However, the rise of generative AI, which can compose music, create images, and write essays resembling human work, prompted EU lawmakers to extend the AI Act to include foundation models. These models, also known as large language models, are trained on vast amounts of data from the internet and enable generative AI systems to create something new, unlike traditional AI that follows predetermined rules.

The recent turmoil at OpenAI, a prominent AI company, further highlighted the risks associated with allowing dominant AI companies to self-regulate. European Commissioner Thierry Breton pointed out that companies like OpenAI prioritize their own interests over the public interest. Surprisingly, France, Germany, and Italy resisted government regulations for these AI systems, advocating for self-regulation instead. This change of stance was seen as an attempt to support homegrown generative AI players and prevent domination of the AI ecosystem by US companies.

A group of influential computer scientists published an open letter expressing their concern that weakening the AI Act would be a "historic failure." Google's top legal officer, Kent Walker, emphasized the importance of regulating AI effectively, stating that the focus should be on the best regulations rather than simply being the first to implement them.

The regulation of foundation models has proven to be a challenging issue for EU negotiators, as it goes against the risk-based approach of the AI Act, which focuses on specific uses of AI. However, regulations are necessary to ensure accountability throughout the AI ecosystem. OpenAI CEO Sam Altman suggested the establishment of a US or global agency to license the most powerful AI systems.

EU negotiators still need to resolve other contentious points, including a proposal to ban real-time public facial recognition entirely. The outcome of these negotiations will determine the future of AI regulation in the European Union and potentially impact global AI governance.