Regulatory Framework for AI-powered Conversational Agents like Chat-GPT Needs Establishment within EU - Regulation Initiative for Managing AI Models like Chat-GPT and Others by the EU
The European Union (EU) has published a voluntary Code of Conduct for Artificial Intelligence (AI) providers, specifically the General-Purpose AI Code of Practice, which aims to assist companies in complying with the forthcoming AI Act. The Code, published in July 2025, is viewed by the EU Commission as an important tool for companies to transition to the new European regulatory framework.
The Code is designed to be a practical tool for industry compliance, prepared by independent experts through a multi-stakeholder process. It applies to all providers of general-purpose AI models, with enhanced requirements applying to powerful AI models with potential risks, such as those that could aid in developing new chemical or biological weapons or where there's a risk of losing control over the technology.
The Code focuses on three key areas: transparency, copyright, and system risks. All providers are addressed in the areas of transparency and copyright, while only providers of the most advanced models considered to pose systemic risks are addressed in the area of system risks.
In terms of transparency, providers must complete a user-friendly Model Documentation Form to detail necessary information for compliance with the AI Act’s transparency obligations. Sufficient transparency is also ensured regarding the operation, limitations, and data sources of general-purpose AI models.
Regarding copyright, providers must establish and implement a policy to comply with EU copyright law, including measures to respect and acknowledge copyrighted materials used in training and operation. The Code offers practical guidance for meeting these obligations, such as documentation and rights management practices.
For system risks, only a small subset of providers of the most advanced models that present systemic risks are addressed. Providers must implement adequate safety and security protocols, as outlined in the AI Act’s obligations for models with systemic risk under Article 55. Providers of advanced models must also be able to demonstrate that they have established risk management and mitigation practices.
Signatories of the Code benefit from reduced administrative burdens and increased legal certainty by demonstrating compliance through the Code. The Code is under assessment by Member States and the Commission, with additional Commission guidelines on key concepts expected in July 2025. Providers who voluntarily sign the Code can use adherence as a means to show compliance with the AI Act’s legal obligations.
Websites that have opted out of automated scraping are to be considered in the implementation of these guidelines. Providers that do not adopt the code will have to develop their own approach to demonstrate compliance, which may involve greater effort. The EU Commission has provided practical guidelines for AI providers to implement EU copyright requirements and expects providers to set up contact points for rights holders.
In conclusion, the EU Code of Conduct for AI provides a practical framework for industry to meet the AI Act’s expectations while offering flexibility and legal clarity to providers of general-purpose AI models. The Code is not mandatory but has a signal effect, with the EU Commission seeing it as a significant tool to help companies comply with the new European regulatory framework for AI.
- The Code of Conduct for AI, which is a practical tool for industry compliance, includes an employment policy for providers, detailing obligations for establishing and implementing a copyright policy to respect and acknowledge copyrighted materials used in training and operation.
- Under the area of system risks in the AI Code of Conduct, a specific employment policy is outlined for providers of advanced models that present systemic risks, requiring them to implement adequate safety and security protocols and to demonstrate established risk management and mitigation practices.
