The European Council and the European Parliament have achieved a provisional agreement on the Artificial Intelligence (“AI”) Act, establishing the world’s first set of rules for AI. The draft regulation aims to guarantee the safety and adherence to fundamental rights and values of the European Union (“EU”) in AI systems deployed within the European market and used across the EU. This milestone proposal signals a historic achievement and clarifies the responsibilities and roles of various actors in AI development and usage.

IN DEPTH

Key Elements of the AI Act Provisional Agreement

Essential elements of the provisional agreement include rules on high-impact general-purpose AI models (such as GPT-4 from OpenAI, Gemini from Google or Llama 2 from Meta), a revised governance system with EU-level enforcement powers, an expanded list of prohibitions, and enhanced protection of rights through mandatory fundamental rights impact assessments for deployers of high-risk AI systems. The agreement adopts a risk-based approach, ensuring that regulations align with the potential harm AI systems may cause. Stricter rules are set for high-impact general-purpose AI models, with severity increasing based on the perceived risk, maintaining a delicate balance between innovation promotion and risk mitigation.

Prohibitions, Exceptions and Law Enforcement Safeguards

The agreement outlines prohibitions on AI practices deemed too risky, including cognitive behavioural manipulation, untargeted scraping of facial images, emotion recognition in workplaces and educational institutions, social scoring, biometric categorisation etc. Exceptions are outlined for law enforcement purposes, subject to specific safeguards, including emergency procedures for urgent situations. Strict criteria are set for real-time remote biometric identification systems in publicly accessible spaces, ensuring fundamental rights protection against potential misuse.

EU’s AI Act about to influence International Regulations

The AI Act is positioned as a flagship legislative initiative that could establish a global standard for AI regulation, much like the General Data Protection Regulation (“GDPR”) influenced international data protection regulations. This landmark agreement is set to shape the European approach to tech regulation on the world stage. To ensure accountability, the agreement introduces fines for violations, set as a percentage of the offending company’s global annual turnover. Proportionate caps are established for Small and Medium Entrepreneurs (“SMEs”) and start-ups. Transparency measures include fundamental rights impact assessments before deploying high-risk AI systems and increased transparency in the use of such systems.

Innovation Support: EU’s AI Regulatory Sandboxes and Measures for Small Businesses

In support of innovation, the agreement modifies provisions to create an innovation-friendly legal framework. AI regulatory sandboxes are introduced for testing and validating innovative AI systems, allowing real-world testing. Actions are outlined to support smaller companies, with limited derogations to ease administrative burdens.

Next steps

The provisional agreement will undergo further technical discussions before formal adoption. The AI Act is expected to come into force two years after its entry into force, marking a significant step towards a comprehensive regulatory framework for AI in the EU.

[View source.]