Europe has reached a significant milestone with a provisional deal on groundbreaking AI rules encompassing artificial intelligence in governmental settings, notably in biometric surveillance, and regulating AI systems such as ChatGPT.
Path to First Major AI Legislation:
This deal positions the EU as a front-runner in enacting comprehensive laws governing AI. Following extensive negotiations spanning nearly 15 hours, the agreement is a crucial step towards regulating AI systems within the EU.
The agreement entails strict obligations for foundation models like ChatGPT and general-purpose AI systems.
They must comply with transparency requirements before entering the market, encompassing technical documentation, EU copyright adherence, and disseminating detailed training content summaries.
Governance and Restrictions:
High-impact foundation models and GP AIs with systemic risks face stringent measures, including model evaluations, systemic risk assessments, adversarial testing, incident reporting, cybersecurity assurance, and energy efficiency disclosures.
Biometric Surveillance and Privacy:
Real-time biometric surveillance in public spaces is limited to specific cases, such as addressing victims of particular crimes, preventing imminent threats like terrorist attacks, and pursuing suspects in severe criminal cases.
The legislation prohibits certain practices, including cognitive behavioral manipulation and unauthorized scraping of facial images or social scoring.
Consumers are granted the right to lodge complaints and receive explanations. Violations of these regulations may incur fines ranging from 7.5 million euros or 1.5% of turnover to 35 million euros or 7% of global turnover.
Criticisms and Opposition:
While hailed as a historical move, criticisms arise. Business groups like DigitalEurope view these regulations as an added burden for companies, especially concerning the regulation of foundation models.
Privacy rights advocates, including European Digital Rights, express concerns about legalized live facial recognition and other aspects of biometric surveillance.
Global Implications and Future Prospects:
The legislation’s anticipated implementation early next year could be a benchmark for other governments seeking to balance AI’s advantages with regulatory frameworks.
This legislation might offer an alternative approach, different from the US’s lighter regulation and China’s interim rules.
