The text will still undergo minor adjustments at the technical level before a committee vote on May 11, but it is expected to go to a plenary vote in mid-June.
Dealing with General Purpose AI, a sub-category of AI that includes ChatGPT, has been a topic of heated debate. The MEPs confirmed previous proposals to put stricter obligations on foundation models. Generative AI models would have to be designed and developed according to EU law and fundamental rights, including freedom of expression.
The MEPs debated what type of AI applications to ban due to their unacceptable risk. AI-powered tools for general monitoring of interpersonal communications were proposed to be prohibited but were dropped following opposition from the European People’s Party (EPP). The EPP, with a strong law enforcement faction, agreed not to table ‘key’ votes that might threaten its support for the entire file. The use of emotion recognition AI-powered software is banned in law enforcement, border management, workplace, and education.
The high-risk classification automatically categorizes AI solutions under critical areas and use cases as high-risk, requiring providers to comply with a stricter regime, including requirements on risk management, transparency, and data governance. MEPs added an extra layer, meaning that AI models falling under Annex III’s categories would only be deemed at high risk if they posed a significant risk of harm to the health, safety, or fundamental rights.
MEPs included extra safeguards for the process whereby the providers of high-risk AI models can process sensitive data such as sexual orientation or religious beliefs to detect negative biases. High-risk AI systems will have to keep records of their environmental footprint, and foundation models will have to comply with European environmental standards.
Ahead of Wednesday’s meeting, the office of the co-rapporteurs circulated a proposal for general principles that would apply to all AI models. These principles include human agency and oversight, technical robustness and safety, privacy and data governance, transparency, social and environmental well-being, diversity, non-discrimination, and fairness.