EU Regulations Call for Transparency from ChatGPT and Similar AI Entities
Tomorrow marks a significant milestone as providers of General-Purpose AI (GPAI) models, such as ChatGPT and Google's AI Gemini, will have to comply with new regulations set by the EU AI Act.
The EU AI Act, adopted in May 2024, includes specific intellectual property (IP) protection measures for AI model providers. These measures are primarily outlined in Article 53 of the Act.
Under the new rules, GPAI model providers will be required to document technical information about their AI models and provide it upon request to the EU AI Office and national authorities. This move aims to increase transparency around model development and training data.
Moreover, providers must implement a policy to comply with Union copyright and related rights laws. This means ensuring their AI models respect copyright regimes in the EU, including how training data is sourced and used.
To aid in fulfilling these obligations, the European Commission has published the General-Purpose AI Code of Practice. This offers practical guidance for GPAI model providers, including chapters on transparency, copyright compliance, and safety/security. The copyright chapter provides guidance on creating robust copyright compliance policies and technical measures to help respect IP rights under the AI Act.
The IP-related provisions of the AI Act aim to facilitate rights holders' ability to exercise and enforce their copyright and related rights against unlicensed or unauthorized use by AI models. However, some creative industry groups have criticized the current implementation as inadequate, arguing it fails to fully safeguard creators’ rights as originally intended by the legislation.
Operators of these AI models will now have to disclose how their systems work and what data they were trained on. Non-compliance with the AI Act could result in fines of up to 15 million euros or three percent of the company's global annual turnover.
Developers must also specify whether they automatically scraped websites for their training data and what measures they took to protect intellectual property. Google, the developer of the AI Gemini, has announced its intention to sign the voluntary code of conduct.
The new rules apply to AI models with general purposes, which are versatile and can write texts, analyze language, or program. Private individuals can already sue providers based on the AI Act.
The European AI Authority will only enforce the new AI Act rules from August 2026 for new models, with older models being controlled from August 2027. Adherence to the voluntary code of conduct could provide providers with higher legal certainty and reduced administrative burden, according to the Commission's assessment.
The EU's guidelines do not require the naming of specific datasets, domains, or sources. Several national and international alliances of authors, artists, and publishers have criticized that the legislation does not sufficiently protect intellectual property. Google has expressed concern that the AI Act could hinder innovation.
According to the EU's guidelines, there will be a contact point for rights holders at the companies. Particularly powerful AI models that could potentially pose a risk to the public will also have to document safety measures. The new rules aim to strengthen copyright, among other things.
The AI Act does not name specific datasets, domains, or sources. The European Commission has published legal guidelines and a voluntary code of conduct to provide guidance for the industry. These new rules aim to ensure lawful use of copyrighted materials in AI training and outputs, while also promoting transparency and safety in AI development.
Service providers of General-Purpose AI (GPAI) models, such as ChatGPT and Google's AI Gemini, will now be required to adhere to specific technology-driven intellectual property (IP) protection measures outlined in Article 53 of the EU AI Act, as they strive to comply with the new regulations set by the EU AI Act. To aid in this process, the European Commission has published the General-Purpose AI Code of Practice, which offers practical guidance on transparency, copyright compliance, and safety/security, including the creation of robust copyright compliance policies and technical measures to respect IP rights under the AI Act.
