The European Parliament has recently passed a series of crucial amendments to the proposed AI legislation of the bloc. MEPs have ratified a set of requirements for foundational models, which form the backbone of generative AI technologies like OpenAI’s ChatGPT.
The amendment, which gained consensus among MEPs in two committees, outlines mandatory obligations for providers of foundational models to conduct safety checks, apply data governance measures and implement suitable risk mitigations prior to releasing their models to the market. This includes a mandate to assess possible risks to health, safety, fundamental rights, the environment, democracy and the rule of law.
As per the amendment, fundamental model developers must strive to minimize the energy usage and resource depletion of their systems, and it is mandatory for them to register their systems in an EU database, which will come into existence under the AI Act.
Additionally, providers of generative AI technologies, including but not limited to Microsoft, must adhere to the transparency norms, ensuring that users are well-informed that the content was generated by a machine. Notable examples of these systems include AI language models like ChatGPT, as well as AI image generators such as Stable Diffusion, Midjourney, and DALL-E.
The proposed legislation mandates that creators of these AI systems must assess and mitigate potential risks before they are made available, including assessing the environmental impact of training such energy-intensive systems and requiring companies to disclose the use of training data protected by copyright law.
Significant topics being discussed by parliament members involve biometric surveillance. The MEPs have concurred and made modifications that strengthen the defense of fundamental rights in this area. Their current goals are centered on the agreement of the parliament’s negotiating mandate for the AI Act. The next stage will enable the EU’s co-legislative procedure to move forward.
Via The Verge