Wed Nov 20 00:18:50 UTC 2024: ## EU Unveils Draft Code of Conduct for Powerful AI Models
**Brussels, Belgium** – The European Union has released its first draft of a Code of Practice for general-purpose artificial intelligence (GPAI) models, aiming to regulate the burgeoning field of advanced AI and prevent potential harms. The 36-page document, published Thursday, provides guidelines for companies developing these powerful AI systems, offering a blueprint for compliance with the recently enacted AI Act.
The code targets AI models trained using immense computing power (over 10²⁵ FLOPs), encompassing major players like OpenAI, Google, Meta, Anthropic, and Mistral, with the potential for further additions. It focuses on several key areas:
* **Transparency:** Companies must disclose details about their data collection methods, including the web crawlers used for training, addressing copyright concerns.
* **Risk Assessment & Mitigation:** The code mandates rigorous risk assessments to prevent issues like cyberattacks, discrimination, and unintended AI behavior. Companies must implement a Safety and Security Framework (SSF) to manage these risks proportionally.
* **Technical Safeguards:** Measures like protecting model data, implementing failsafe access controls, and continuous risk reassessment are required.
* **Governance:** Internal accountability mechanisms and the engagement of external experts are emphasized.
Failure to comply with the code could result in significant penalties – fines of up to €35 million or 7% of global annual turnover, whichever is greater.
The draft is open for public comment until November 28th via the Futurium platform. The EU aims to finalize the code by May 1, 2025. This initial draft represents a crucial step towards establishing a regulatory framework for advanced AI, balancing innovation with the mitigation of potential risks.
First Piper 