After an exhaustive 72-hour debate, European Union lawmakers achieved a landmark agreement on Friday regarding the expansive AI Act safety development bill. This legislation, reported by The Washington Post as the broadest-ranging and most far-reaching of its kind to date, marks a significant milestone. While specific details of the deal were not immediately available, Dragoș Tudorache, a Romanian lawmaker who co-led the AI Act negotiation, emphasized its potential impact.
Tudorache stated to The Washington Post, “This legislation will set a standard, serving as a model for many other jurisdictions. We must exercise an extra duty of care in its drafting, considering its influential role for many others in the future.”
— Thierry Breton (@ThierryBreton) December 8, 2023
The proposed regulations aim to govern the development and distribution of future machine learning models within the trade bloc, influencing their applications across various sectors such as education, employment, and healthcare. The categorization of AI development into four tiers—minimal, limited, high, and banned—reflects the societal risk associated with each.
Banned applications encompass activities that violate user consent, target protected social groups, or involve real-time biometric tracking, like facial recognition. High-risk uses include AI intended for safety components in products or designated for critical sectors such as infrastructure, education, legal matters, and employee hiring. Chatbots like ChatGPT, Bard, and Bing fall under the “limited risk” category.
Dr. Brandie Nonnecke, Director of the CITRIS Policy Lab at UC Berkeley, highlighted the European Commission’s bold approach in addressing emerging technology, drawing parallels to their proactive stance on data privacy with GDPR. She noted the proposed regulation’s unique focus on a risk-based approach, similar to what has been suggested in Canada’s proposed AI regulatory framework.
Recent disruptions in the ongoing negotiations over proposed rules were instigated by France, Germany, and Italy. These countries were obstructing talks related to guidelines for the development of Foundational Models by EU member nations. Foundational Models, like OpenAI’s GPT-4, serve as generalized AIs from which more specialized applications can be refined. Concerns were raised by this trio of nations, fearing that stringent EU regulations on generative AI models could impede member nations’ competitive development efforts.
The European Commission (EC) had previously tackled the challenges associated with managing emerging AI technologies through various initiatives. In 2018, the EC released the first European Strategy on AI and Coordinated Plan on AI, followed by the Guidelines for Trustworthy AI in 2019. Subsequent efforts included a White Paper on AI and a Report on the safety and liability implications of Artificial Intelligence, the Internet of Things, and robotics in the following year.
In its draft AI regulations, the European Commission emphasized that artificial intelligence should be viewed as a tool serving people, aiming to enhance human well-being. The regulations proposed for artificial intelligence in the Union market or affecting Union citizens prioritize a human-centric approach, ensuring that people can trust that the technology is used in a safe and legally compliant manner, respecting fundamental rights.
Simultaneously, the directive emphasized the need for rules governing artificial intelligence to be balanced, proportionate, and avoid unnecessary constraints or impediments to technological progress. The statement highlighted the challenge of foreseeing all potential future uses or applications of artificial intelligence, given its pervasive presence in various aspects of people’s daily lives.
In more recent developments, the European Commission (EC) has initiated collaborations with industry stakeholders on a voluntary basis to formulate internal regulations. These rules aim to establish a common framework for companies and regulators, even before formal AI regulations take effect. Thierry Breton, the EC industry chief, stated in a May release that he and Google CEO Sundar Pichai agreed on the urgency of not waiting for AI regulation to become applicable. Instead, they advocated working with all AI developers to create a voluntary AI pact ahead of the legal deadline. Similar discussions with U.S.-based corporations have also been initiated by the EC.
The situation is currently evolving, with further developments anticipated.