Question to the Department for Science, Innovation & Technology:

To ask the Secretary of State for Science, Innovation and Technology, with reference to the oral evidence of the Prime Minister to the Liaison Committee on 4 July 2023, HC1602, what steps her Department is taking to implement the (a) reporting and (b) licensing of large training runs for artificial intelligence.

It is clear that the right guardrails must be in place to manage the risks AI poses. Our proportionate regulatory framework, underpinned by a set of principles and supported by tools like AI assurance techniques and technical standards, sets out a responsible approach to AI innovation. The Government also made it clear in our AI regulation white paper that our approach must be adaptable. As we now look to implement the new regulatory regime, we are considering how the framework will apply to the various actors in the AI development and deployment lifecycle, with a particular focus on foundation models.

This is supported by the £100 million Foundation Model Taskforce led by Ian Hogarth, as well as a new central risk function and international leadership on AI safety through the AI summit.

Our proposed central functions, including risk analysis, horizon scanning, and monitoring and evaluation, will keep the wider landscape under constant review to inform policy. They will capture emerging risks, including risks arising from increasingly powerful foundation models. The Foundation Model Taskforce will meanwhile advance vital safety research, laying the groundwork for the safe adoption of AI across the UK economy, ensuring we are at the forefront of this pivotal technology.

The UK will host the first major global summit on AI safety this autumn. The Summit will bring together key countries, as well as leading technology companies and researchers, to drive targeted, rapid international action to guarantee safety and security at the frontier of this technology.

As the Prime Minister told the Liaison Committee, the Government continues to analyse a range of safety features and guard rails that we could put in place. These could include new measures for the reporting and licensing of large training runs. But it is vital that the effectiveness of any proposals are rigorously evaluated before they are implemented. This is why we welcome the wide range of stakeholders that provided insights to our consultation on the AI regulation white paper. We are currently considering all evidence sent to the consultation and we will provide an update through the Government's response later in the year.