The UK has deliberately chosen a different path from the EU in regulating AI. While the EU AI Act (Regulation 2024/1689) creates a comprehensive, horizontal regulatory framework with risk-based classification (unacceptable, high, limited, and minimal risk), the UK's white paper "A Pro-Innovation Approach to AI Regulation" (March 2023) proposes a principles-based, sector-specific approach. Five cross-cutting principles — safety/security/robustness, transparency/explainability, fairness, accountability/governance, and contestability/redress — are to be implemented by existing regulators (FCA, ICO, Ofcom, CMA, MHRA, etc.) within their existing mandates. The AI Safety Institute (AISI), established at Bletchley Park following the AI Safety Summit in November 2023, focuses on evaluating frontier AI models for safety risks — catastrophic, societal, and systemic. The AISI conducts pre-deployment testing of advanced AI models, develops safety evaluation frameworks, and publishes research on AI risk. The UK has signalled it may introduce legislation to give regulators formal AI responsibilities (a potential "AI Bill"), but the timeline remains uncertain. In the interim, regulators are acting under existing powers: the ICO has published guidance on AI and data protection (including a risk toolkit), the FCA has issued expectations for firms using AI in financial services, and the EHRC has signalled that algorithmic discrimination under the Equality Act 2010 is an enforcement priority. For businesses, the practical implication is navigating a patchwork of sector-specific expectations rather than a single horizontal framework.
AI Safety Institute (AISI)
A UK Government body (originally the Frontier AI Taskforce) established in November 2023, headquartered at Bletchley Park. Its mission is to advance the world's knowledge of AI safety by: evaluating advanced AI systems before and after deployment, conducting fundamental research on AI safety, and facilitating information exchange between governments, developers, and researchers. The AISI has agreements with leading AI developers (including Anthropic, Google DeepMind, Meta, Microsoft, and OpenAI) to conduct safety evaluations of frontier models.