Question to the Department for Science, Innovation & Technology:

To ask the Secretary of State for Science, Innovation and Technology, what transparency conditions are currently required when government departments procure AI systems from private companies; and what mechanisms are in place to ensure public sector bodies can explain AI-driven decisions to citizens when the underlying models are proprietary.

Since February 2024, all government departments and arm’s-length bodies must comply with the Algorithmic Transparency Recording Standard (ATRS), which mandates publishing details on algorithmic tools, including decision-making processes, human oversight, technical specifications, and risk assessments. Suppliers are required to provide sufficient information for transparency records, with exemptions balancing commercial sensitivities. Over 36 ATRS records have been published to date.

The AI Knowledge Hub further enhances transparency by sharing open-source code, problem statements, and performance metrics.

Additionally, the Open Source AI Fellowship promotes explainability through publicly inspectable models. These measures enable government to explain AI-driven decisions while maintaining accountability.