The European AI Office on 14th November 2024 unveiled the first draft of the General-Purpose AI Code of Practice, marking a critical step in regulating AI development and deployment under the AI Act. This draft was developed collaboratively by four thematic working groups composed of independent experts. It addresses transparency, systemic risk management, technical safeguards, and governance for providers of general-purpose AI models.
Key Highlights of the Draft
- Transparency and Copyright Compliance: Providers are required to maintain detailed documentation about their AI models, including training processes and compliance with copyright laws.
- Managing System Risks: Advanced models that pose systemic risks must adopt strict safety and security frameworks. This includes taxonomy for identifying risks such as cyber vulnerabilities, misinformation, and large-scale discrimination.
- Governance and Accountability:Companies must implement governance mechanisms to ensure systemic risk assessments, independent evaluations, and compliance with safety standards.
- Adaptability: The Code emphasizes flexibility, ensuring rules evolve with technological advancements while maintaining a focus on protecting fundamental rights.
Engagement and Feedback
Stakeholders will convene next week for a plenary discussion with nearly 1,000 participants, including representatives from EU Member States, academia, and industry leaders. Feedback is also being collected via the Futurium platform until November 28, 2024. Future drafts will incorporate this feedback, with the final version expected in April 2025.
Significance
The Code is positioned as a vital bridge between current practices and future EU standards, offering clarity for AI developers while promoting transparency and trustworthiness. It is expected to serve as a global benchmark for AI governance, fostering collaboration across industries and countries. The initiative highlights Europe’s leadership in setting ethical and technical standards for AI while ensuring inclusivity and adaptability in addressing systemic risks posed by advanced AI systems.
For more details, Read the draft here: First_Draft_GeneralPurpose_AI_Code_of_Practice__sFt8VTOzxWsmJhIfdMVYPLM42C0_109946.pdf