UK AISI to Gain Access to Advanced AI Models Before Public Release

Shortly after its founding in November 2023, the UK AI Safety Institute (AISI) began assessing the risks of frontier AI models before they were released to the public.

UK and US Discuss Access to AI Safety Testing

It was reported in September, external that the White House had asked some major US research institutions not to share models with research institutions until the US had completed its own testing.

The move was seen at the time as a setback for AISI, which has been supported by the UK government and its partners to promote open, constructive and collaborative AI testing.

Microsoft president Brad Smith recently told the BBC that the institute had helped establish the UK as the country best suited to drive global AI regulation.

Narayan said on Monday that the UK government had been working with the US in recent weeks to “ensure that the UK AISI has access to the most powerful AI models before they are published.”

UK Considers Stronger AI Incident Protections

Narayan also said the UK would consider introducing stronger protections, particularly for identifying and reporting serious AI incidents, through the Cybersecurity and Resilience Bill.

The announcement follows a series of incidents in which AI agents from major companies, including OpenAI and Anthropic, acted outside human control and hacked external systems.

Calls for Greater AI Safety Regulation

Executives at US-based AI companies such as Anthropic, OpenAI and Google DeepMind support the idea of slowing AI development.

However, there are currently no global safety regulations governing AI companies and their technologies.

Meanwhile, President Trump has called for industry self-regulation to prevent US companies from losing ground to Chinese rivals.