Europe has reached a provisional agreement on landmark European Union rules governing the use of artificial intelligence, including governments’ use of AI in biometric surveillance and how to regulate AI systems like ChatGPT.
With the political agreement, the EU moves closer to becoming the first major world power to pass AI legislation. The agreement by EU countries and European Parliament members came after nearly 15 hours of negotiations that followed a nearly 24-hour debate.
The two sides are expected to work out details in the coming days, which could alter the final legislation’s shape.
“Europe has positioned itself as a pioneer, understanding the importance of its role as a global standard setter. This is yes, I believe, a historical day,” European Commissioner Thierry Breton said duirng a press conference.
The agreement requires foundation models such as ChatGPT and general purpose AI systems (GPAI) to meet transparency requirements before they are released to the public. These include creating technical documentation, adhering to EU copyright law and disseminating detailed summaries of training content.
High-impact foundation models with systemic risk must conduct model evaluations, assess and mitigate systemic risks, conduct adversarial testing, report serious incidents to the European Commission, ensure cybersecurity and report on energy efficiency.
To comply with the new regulation, GPAIs with systemic risk may rely on codes of practice.
Governments can only use real-time biometric surveillance in public spaces to identify victims of specific crimes, to prevent genuine, current, or foreseeable threats such as terrorist attacks and to search for people suspected of the most serious crimes.
The agreement prohibits cognitive behavioral manipulation, untargeted scraping of facial images from the internet or CCTV footage, social scoring and biometric categorization systems used to infer political, religious, philosophical beliefs, sexual orientation and race.
Consumers would be able to file complaints and receive meaningful explanations, with fines ranging from 7.5 million euros ($8.1 million) or 1.5% of global turnover to 35 million euros or 7% of global turnover.
Europe’s ambitious AI rules come as companies like OpenAI, in which Microsoft (MSFT.O) is a shareholder, continue to find new applications for their technology, eliciting both praise and criticism. Alphabet (GOOGL.O), the parent company of Google, launched Gemini, a new AI model to compete with OpenAI.
The EU law could serve as a model for other governments, as an alternative to the US’ light-touch approach and China’s interim rules.
