According to Cointelegraph: New rules governing the use of artificial intelligence (AI) have seen a provisional agreement reached by negotiators for the European Parliament and Council. Components of the agreement, made on Friday, December 8th, cover the government use of AI in biometric surveillance, regulating AI systems like ChatGPT, and transparency protocols to follow before market entry. This includes criteria around technical documents, adherence to European Union copyright laws, and the sharing of training content summaries.
The EU's goal is to become the first supranational authority to enact AI-specific laws, with the dual aims of benefiting from its use and mitigating associated risks. The deliberations leading to the agreement followed a nearly 24-hour debate and a following 15-hour negotiation period.
Under the agreement, AI models that might pose systemic risks or have a significant impact must evaluate and address these risks. The models must also undergo adversarial testing to assess system resilience, report incidents to the European Commission, ensure cybersecurity and disclose energy efficiency.
The agreement also imposes rules on general-purpose artificial intelligence. The new regulations grant consumers the right to lodge complaints and seek explanations, while banning practices like real-time biometric surveillance, cognitive behavioral manipulation, image scraping, social scoring, and biometric systems that infer personal details. There are also penalties for non-compliance, ranging from 7.5 million euros ($8.1 million) or 1.5% of turnover to 35 million euros ($37.7 million) or 7% of global turnover.
Following formal ratification by the parliament and council, the agreed text will be enacted as EU-wide law, subject to an approving vote by the parliament’s internal market and civil liberties committees at an upcoming meeting.