The European Union’s AI Act has advanced to the trilogue phase following nearly two years of development. This legislation sets forth specific requirements for generative AI tools, such as ChatGPT, mandating that they reveal any copyrighted material utilized in their creation.
The primary objective of the act is to establish a framework for regulating emerging AI technologies, which have gained considerable momentum recently. Notable developments include Alphabet’s formation of a new AI-focused division and the UK government’s commitment to leading the AI sector.
The proposed legislation categorizes AI tools based on their associated risk levels, ranging from minimal to high and unacceptable. Key areas of concern encompass biometric surveillance, the dissemination of misinformation, and potential discriminatory language.
While high-risk tools are not outright banned under this draft, they will be subject to stringent transparency requirements. Creators of generative AI tools, such as image generators like Midjourney, must also adhere to these transparency standards by disclosing the materials used in their development.
Recent reports indicate that this specific provision was introduced in response to discussions within the last two weeks. Some committee members initially proposed an outright ban on the use of copyrighted material for training generative AI models, but this suggestion was ultimately replaced with a focus on transparency.
Svenja Hahn, a member of the European Parliament, remarked on the compromise reached: “Despite opposing views advocating for increased surveillance or excessive regulation, Parliament has struck a balance that effectively regulates AI, safeguards citizens’ rights, and promotes innovation and economic growth.”