The US and the UK signed a Memorandum of Understanding (MOU) to develop standards for AI Safety on April 1.

The agreement aims to develop and iterate “robust suites of evaluations” for AI models, systems, and agents through coordinated scientific approaches.

 

The two countries will specifically conduct at least one joint testing exercise on a publicly accessible AI model. They also intend to exchange personnel between their previously established AI Safety Institutes. The partnership takes effect immediately.

The US Department of Commerce and the UK’s Department for Science, Innovation and Technology announced the partnership in identical press releases.

Safe AI development

Commenting on the new international agreement, US Commerce Secretary Gina Raimondo called AI the “defining technology of our generation” and said that the partnership will address all risks related to the technology, from national security risks to societal risks.

Raimondo described the partnership’s proactive nature, stating:

“Our partnership makes clear that we aren’t running away from these concerns – we’re running at them.”

UK Technology Secretary Michelle Donelan called the agreement a “landmark moment” and called the pursuit of safe AI development a global issue.

 

In a separate interview with the Financial Times, Donelan emphasized the urgency of the partnership, considering the upcoming release of new AI models over the next year. She added that these models’ full capabilities are unknown, but they are expected to be “game-changing.”

Donelan highlighted that many top AI companies are US-based, which necessitates information exchange with the government.

Other AI developments

The current agreement builds on commitments to AI made during the UK’s AI Safety Summit, which was held in November 2023. The broader commitment to AI safety is part of the Bletchley Declaration, a policy paper from countries that attended the summit.

The news comes shortly after the completion of goals in the Biden Administration’s executive order in the US, which culminated in orders to appoint chief AI officers across all federal agencies on March 28.

The UK set out a local approach to regulation in February.