Question to the Ministry of Defence:

To ask the Secretary of State for Defence, for which roles his Department has considered using non‑deterministic AI systems; and what mechanisms he plans to put in place to (a) monitor and (b) mitigate associated risks.

The Ministry of Defence intends to exploit AI across a broad spectrum of use cases from the frontline to the back office to make all parts of Defence more efficient and effective. This may include the use of ‘non-deterministic’ AI systems (such as Large Language Models). Joint Service Publication 936 on Dependable AI provides a robust framework to address AI-related risks around human control and unpredictability as part of our existing risk management process. This ensures that our legal, ethical and safety obligations are always fulfilled for all AI use cases. We expect that use cases such as training simulations and operational planning exercises will require robust assurance so that outcomes remain valid and repeatable.