Friday sees Europe clinch provisional deal on groundbreaking EU rules covering AI, encompassing government biometric surveillance and ChatGPT-like system regulation.EU nears landmark status as first major AI regulator. Friday's political accord, after nearly 15 hours of negotiations, follows a marathon 24-hour debate, positioning the EU as a pioneer in AI law.
In the coming days, both sides will finalize details, potentially altering the final legislation's contours.
European Commissioner Thierry Breton declares a historic day, emphasizing Europe's pioneering role as a global standard setter in AI.The agreement mandates foundational models like ChatGPT and general-purpose AI systems (GPAI) to meet transparency obligations pre-market release. This involves crafting technical documentation, adhering to EU copyright law, and providing comprehensive summaries of the training content.
Foundation models with systemic risk must undergo evaluations, mitigate risks, conduct adversarial testing, report incidents to the European Commission, ensure cybersecurity, and disclose energy efficiency.GPAIs carrying systemic risk can adhere to the new regulation through reliance on codes of practice.
Governments are restricted to employing real-time biometric surveillance in public spaces for specific purposes, including aiding victims of certain crimes, preventing genuine and imminent threats (such as terrorist attacks), and locating individuals suspected of the most serious offenses.
The agreement prohibits cognitive behavioral manipulation, indiscriminate scraping of facial images from the internet or CCTV footage, social scoring, and biometric categorization systems for deducing political, religious, philosophical beliefs, sexual orientation, and race.Consumers retain the right to file complaints and receive substantive explanations. Fines for breaches range from 7.5 million euros ($8.1 million) or 1.5% of turnover to 35 million euros or 7% of global turnover.
Business group DigitalEurope criticized the rules, viewing them as an additional burden for companies, compounded by recent legislation.
"We have a deal, but at what cost? While we fully endorsed a risk-based approach tied to the uses of AI, rather than the technology itself, the last-minute effort to regulate foundation models has reversed our stance," remarked Cecilia Bonefeld-Dahl, Director General.
Privacy rights group European Digital Rights was equally critical.
"It's challenging to find enthusiasm for a law that, for the first time in the EU, has moved towards legalizing live public facial recognition across the bloc," commented Ella Jakubowska, Senior Policy Advisor.
"While the Parliament made efforts to minimize the impact, the comprehensive package on biometric surveillance and profiling is, at best, lukewarm,"The legislation is anticipated to take effect early next year, pending formal ratification by both sides, and is slated to be applicable two years thereafter.
Governments globally strive to strike a balance between harnessing the benefits of technology capable of human-like conversations, answering questions, and writing computer code, while concurrently implementing necessary safeguards.
Europe's ambitious AI regulations emerge amidst ongoing advancements by companies like OpenAI, backed by investors such as Microsoft (MSFT.O), as they explore new applications for their technology, eliciting both praise and apprehension. On Thursday, Google's parent company, Alphabet (GOOGL.O), introduced a new AI model, Gemini, to compete with OpenAI.
The EU law might serve as a model for other governments, offering an alternative to the United States' hands-off approach and China's interim rules.
($1 = 0.9293 euros)
