Policy briefing - An EU AI Act that works for people and society

What role are different types of AI systems likely to play in our future economy and society? The AI Safety Summit is premised on the notion that the impact of AI on our society and economy will be transformational.3 In some ways, it already is: AI is being deployed in important areas of scientific discovery such as genomics,4 and across important societal challenges such as climate change adaptat. [...] A challenge for Government – particularly where the stakes are high – is that many of the risks of AI are not related to the ‘capabilities’ of a base model but rather the functionality of a specific AI system in an applied context.16 There is already considerable evidence of AI use and piloting within the UK public sector, with recent reports indicating government applications across benefit decis. [...] Examples of this include: • Medicines and medical devices: The UK’s Medicines and Healthcare products Regulatory Agency’s (MHRA) goal is to ‘ensure the safety and effectiveness of medical products within the UK [...] we aim to safeguard patient well-being and maintain public trust in the healthcare sector’. [...] Many of these gaps relate to the capacity of regulatory actors to ‘reach’ different contexts in which AI will be used, and to shape the practices of actors further up the value chain, such as developers and hosts of AI systems, particularly for foundation models. [...] Mission critical: Lessons from relevant sectors for AI safety 15 We’ll be exploring in more depth the questions raised in this briefing, such as: • the types of objectives and public benefits that these regulatory regimes aim to achieve • the mechanisms put in place at different stages of the value chain to achieve these objectives and benefits • the distribution of liability and compliance burden.