Question to the Department for Science, Innovation & Technology:

To ask the Secretary of State for Science, Innovation and Technology, whether civil servants receive (a) training and (b) guidance on managing the risks of non-deterministic outputs in AI-assisted decision-making.

DSIT has provided civil servants with both training and guidance on managing the risks of AI, including non-deterministic models such as LLMs. Training includes over 70 free courses on AI fundamentals, ethics, machine learning and generative AI, available through Civil Service Learning and the Government Campus. These emphasise critical evaluation of AI outputs and awareness of risks like bias and hallucination. The AI Playbook for the UK Government outlines principles for managing risks, accuracy, bias, hallucinations, and security threats. Civil servants are advised to apply human oversight and robust testing to ensure responsible use of AI and always adhere to the Model of Responsible Innovation.