In a move signaling Europe's leadership in the digital age, the European Union has reached a provisional deal on pioneering rules to govern the use of artificial intelligence (AI). The agreement, achieved after exhaustive negotiations, positions the EU as the world's first major power to enact comprehensive laws in this rapidly evolving field.
The deal, which came to fruition following nearly 15 hours of negotiations and a marathon 24-hour debate, outlines stringent regulations governing AI applications, including biometric surveillance by governments and the regulation of AI systems like ChatGPT. European Commissioner Thierry Breton, at a press conference, hailed this agreement as a historic milestone, marking Europe as a global standard setter in AI governance.
One of the critical aspects of the agreement is the requirement for foundation models, such as ChatGPT, to adhere to transparency obligations before market deployment. This includes creating technical documentation, complying with EU copyright law, and providing detailed summaries of training content.
High-impact foundation models, identified as posing systemic risk, will be subject to rigorous evaluations, including model assessments, systemic risk mitigation, adversarial testing, and reporting on serious incidents and energy efficiency.
The agreement also delineates the limited circumstances under which governments can use real-time biometric surveillance in public spaces, strictly in cases involving serious crimes or genuine threats such as terrorist attacks. It further prohibits practices like cognitive behavioral manipulation, untargeted scraping of facial images from the internet, social scoring, and biometric categorization systems used to infer sensitive personal characteristics.
The deal has not been without its critics. Business group DigitalEurope expressed concerns about the additional burdens the rules place on companies. Conversely, privacy rights group European Digital Rights criticized the legislation for not going far enough in curbing public facial recognition and profiling.
The agreement also sets significant fines for violations, ranging from 7.5 million euros or 1.5% of turnover to 35 million euros or 7% of global turnover. Consumers affected by AI violations will have the right to launch complaints and receive detailed explanations.
This landmark legislation, expected to take effect early next year and apply two years after formal ratification, could serve as a template for other global powers. It offers an alternative to the United States' light-touch approach and China's interim rules, potentially shaping global AI governance norms.
The AI Act's inception was originally designed to mitigate dangers from specific AI functions based on their risk level. However, the surge in generative AI's popularity necessitated updates to the proposal. These generative AI systems, like OpenAI's ChatGPT and Google's Bard chatbot, capable of producing human-like text and images, have sparked both admiration and apprehension regarding their potential impacts on privacy, employment, and even existential risks to humanity.
The EU's foray into AI regulation marks a significant step in balancing the benefits of this transformative technology with the need for ethical guardrails. As AI continues to integrate into various facets of life, this regulatory framework could be pivotal in ensuring AI's advancements align with societal values and rights.
