US, UK announce partnership on AI safety, testing


The United States and Britain on Monday announced a new partnership on the science of artificial intelligence security, amid growing concerns about upcoming next-generation versions.

Following commitments announced at the AI ​​Security Summit at Bletchley Park in November, Commerce Secretary Gina Raimondo and British Technology Secretary Michelle Donnellan signed a memorandum of understanding in Washington to jointly develop advanced AI model testing.

“We all know that AI is the defining technology of our generation,” Raimondo said. “This partnership will accelerate the work of both of our institutions on a broader scale to address the risks to our national security concerns and the concerns of our broader society.”

Britain and the United States are among the countries setting up government-led AI safety institutes.

Britain said in October that its institute would investigate and test new types of AI, while the United States said in November it was starting its own security institute to evaluate the risks from so-called frontier AI models And now working with 200 companies and institutions.

Under the formal partnership, Britain and the United States plan to conduct at least one joint testing exercise on a publicly accessible model and are considering exploring personnel exchanges between the institutions. Both are working to develop similar partnerships with other countries to promote AI security.

Also read: US needs new AI safeguards for government use, transparency

“This is the first agreement of its kind anywhere in the world,” Donnellan said. “AI is already an extraordinary force for good in our society, and it has huge potential to tackle some of the world’s biggest challenges, but only if we are able to overcome those risks.”

Generative AI – which can generate text, photos and video in response to open-ended prompts – has sparked excitement as well as fears that it could render some jobs obsolete, influence elections and potentially can overpower humans and have devastating effects.

In a joint interview with reuters On Monday, Raimondo and Donnellan called for urgent joint action to address AI risks.

“The timing is very important because the next set of models are about to be released, which will be much more capable,” Donnellan said. “Our focus is on the areas that we are dividing, conquering and really gaining expertise in.”

Raimondo said she would raise AI issues at the US-EU Trade and Technology Council meeting in Belgium on Thursday.

Reading: Asahi says US, Japan will call for deeper cooperation in AI, semiconductors

Raimondo said the Biden administration plans to announce additional people to its AI team soon. “We’re drawing the entire resources of the U.S. government.”

Both countries plan to share key information on capabilities and risks associated with AI models and systems and technical research on AI security.

In October, Biden signed an executive order aimed at reducing the risks of AI. In January, the Commerce Department said it was proposing to require US cloud companies to determine whether foreign entities are accessing US data centers to train AI models.

Britain said in February it would spend more than 100 million pounds ($125.5 million) to launch nine new research centers and an AI train regulator on the technology.

Raimondo said she was particularly concerned about the threat of AI applied to bioterrorism or nuclear war simulations.

“These are things where the consequences could be devastating and so we really have to have zero tolerance towards some of the models being used in that capacity,” he said.