EU: Commission publishes second draft of GPAI (General Purpose Artificial Intelligence) code of practice

On December 19, 2024, the European Commission published the second draft of the 'General-Purpose AI Code of Practice.' Independent experts present the second draft of the General-Purpose AI Code of Practice, based on the feedback received on the first draft, published on 14 November 2024. 

In addition to the written feedback to be received by January 15, 2025, verbal discussions on the second draft with Chairs and Vice-Chairs are planned in line with the tentative timeline published by the AI Office. The third draft of the Code of Practice can be expected for the week of February 17, 2025.   

Summary high-level principles

Here are 7 bullet points summarizing the high-level principles for drafting the Code: 

• Alignment with EU Values: Commitments must align with EU principles and values, including those enshrined in the Charter of Fundamental Rights. 
• AI Act and International Approaches: Compliance with the AI Act and consideration of international approaches (e.g., standards by AI Safety Institutes) are crucial. 
• Proportionality to Risks: 
  • Commitments and KPIs should be tailored to the level of risk (e.g., higher risk tiers require stricter measures). 
  • Measures and KPIs must be specific, effective, and avoid unnecessary burdens. 
  • Differentiate between risk types, distribution strategies, and deployment contexts. 

• Future-Proofing:  
  • Allow for rapid updates to Measures and KPIs as AI technology evolves.
  • Reference dynamic sources of information (e.g., incident databases, consensus standards). 

• Consideration of Provider Size: 
  • Allow for simplified compliance for SMEs and startups. 
  • Adjust KPIs based on provider size and capacity, especially for those with systemic risk models. 

• Support for a Safe and Trustworthy AI Ecosystem:  
  • Encourage cooperation between stakeholders (e.g., sharing best practices, civil society participation). 
  • Acknowledge the positive impact of open-source models. 

• Innovation in AI Governance:  
  • Encourage providers to innovate in AI safety governance and evidence collection methods. 
  • Recognize and support alternative approaches with equal or superior safety outcomes.

View full document here