The regulators charged with implementing the Government’s high-level AI governance principles should, in their approach to these models and tools, prioritise testing and verifying their outputs, as well seeking to establish—whilst accepting the difficulty of doing so with absolute certainty—how they arrived at them.