European Union negotiators achieved a landmark deal on Friday, establishing the world’s inaugural comprehensive artificial intelligence rules. The breakthrough paves the way for legal oversight of transformative technologies, notably generative AI services such as ChatGPT, which have promised to reshape daily life while simultaneously raising concerns about existential threats to humanity.
After intense negotiations between the European Parliament and the bloc’s 27 member countries, a tentative political agreement for the Artificial Intelligence Act was reached, marking a significant step toward regulating AI technologies. European Commissioner Thierry Breton celebrated the achievement, tweeting “Deal!” just before midnight. The parliamentary committee co-leading the negotiations confirmed the agreement, signaling a significant milestone in AI governance.
Details regarding the specific provisions of the law that will take effect no earlier than 2025 remain limited, as negotiators faced challenges in aligning on controversial points, including the regulation of generative AI and police use of facial recognition surveillance. The closed-door talks, lasting for extended periods, culminated in a compromise that reflects the EU’s commitment to establishing guidelines for emerging AI technologies.
The EU took an early lead in developing AI regulations with the unveiling of its rulebook in 2021. The surge in generative AI advancements prompted a rapid update of the proposal, aiming to address risks related to jobs, privacy, copyright protection, and potential threats to human life. Generative AI systems, exemplified by OpenAI’s ChatGPT, have garnered attention for their ability to produce human-like text, photos, and songs.
As the U.S., U.K., China, and global coalitions like the Group of 7 introduce their own AI regulatory proposals, Europe remains at the forefront. The final version of the AI Act requires approval from the bloc’s 705 lawmakers, expected to be a formality, before EU-wide elections next year.
Originally designed to mitigate risks based on AI functions, the AI Act expanded to include foundation models, the advanced systems underpinning general-purpose AI services. Despite initial challenges, negotiators reached a compromise addressing concerns raised by France, which advocated for self-regulation to support European generative AI companies.
Foundation models, often large language models trained on extensive data sources, are crucial for generative AI’s ability to create new content. The deal mandates additional scrutiny for the most advanced foundation models posing “systemic risks,” including requirements for enhanced disclosure of information such as the computing power used in training.
This groundbreaking agreement signifies a crucial step toward responsible AI governance, balancing technological innovation with safeguards against potential risks and ensuring transparency in the development of AI-powered services.
