Question to the Department for Science, Innovation & Technology:

To ask the Secretary of State for Science, Innovation and Technology, if he will publish guidance on the safe use of non-deterministic AI in public sector decision-making systems.

The AI playbook clearly describes the non-deterministic nature of probabilistic, AI-based predictive analysis and automated decision-making, while prescribing ten key principles for the safe and responsible use of AI. For example, it highlights the non-deterministic and probabilistic nature of AI models—specifically large language models (LLMs)—which can yield inaccurate or hallucinated outputs. Accordingly, it stresses legal and ethical duties, requiring departments to register AI-powered algorithms via the Algorithmic Transparency Recording Standard. It also recommends and emphasises meaningful human oversight, including monitoring, assurance, and user feedback mechanisms, with a risk-aware approach from inception through the entire life cycle of the services.