On Friday, the European Union made a significant stride in regulating artificial intelligence (AI) technologies, including those like OpenAI's ChatGPT, by finalizing a groundbreaking set of rules. This development marks a crucial step in establishing legal oversight over this increasingly prevalent technology.
Negotiations between the European Parliament and its 27 member states were challenging, with key issues such as generative AI, facial recognition, and the regulation of foundational AI models like ChatGPT and Google's Bard at the forefront. These discussions, reported by the Associated Press, revealed significant differences in opinions.
One major point of contention involved AI-driven facial recognition systems, with EU lawmakers considering outright bans due to privacy concerns. This stance conflicted with some member states' desires to use these systems for law enforcement. After an intense 22-hour negotiation session that ended in a deadlock on Thursday, talks resumed on Friday.
The initiative to establish AI regulations began in Brussels in 2021, positioning Europe as a leader in shaping the governance of AI. This move addresses growing apprehensions about AI's potential impact on privacy, employment, and copyright laws.
While the agreement awaits a formal vote by the European Parliament, expected early next year, Italian lawmaker Brando Benifei, a key figure in the negotiations, anticipates this will be a mere formality, according to the Associated Press.
Reuters reports that the new regulations will impose several obligations on foundational AI models like ChatGPT. These include the creation of technical documentation, adherence to EU copyright laws, and the provision of detailed summaries about their training content.
For models posing a high impact and systemic risk, additional measures such as model evaluations, risk mitigation, adversarial training, cybersecurity assurances, energy efficiency reporting, and notifying the European Commission about serious incidents are mandatory.
The deal also allows limited government use of real-time biometric surveillance in public spaces, specifically for identifying crime victims, preventing imminent threats, including terrorism, and locating individuals suspected of serious crimes.
Furthermore, the agreement prohibits practices like cognitive behavioral manipulation, unauthorized collection of facial images, and the use of biometric systems that infer personal details such as religious, political, or philosophical beliefs, race, and sexual orientation.
Comments