The US and UK have signed a landmark agreement on artificial intelligence, making the allies the first countries to formally cooperate on how to test and assess the risks of emerging AI models.
The agreement, signed in Washington on Monday by UK Science Minister Michelle Donelan and US Commerce Secretary Gina Raimondo, sets out how the two governments will share expertise, knowledge and capabilities on the security of artificial intelligence (AI). The agreement is the world’s first bilateral agreement on AI security.
“The year ahead is one where we really need to move fast because the next generation of [AI] models are emerging that could be a complete game changer and we don’t yet know all the capabilities they will offer,” Donelan told the Financial Times (FT).
In particular, the agreement will allow the UK’s new Artificial Intelligence Security Institute (AISI), launched in November, and its US counterpart, which has yet to begin work, to share expertise through the secondment of researchers from both countries. The institutes will also work together on how to independently assess specific AI models developed by organisations such as OpenAI and Google.
Relationship between intelligence agencies to be modelled
The partnership is modelled on that between the UK’s Government Communications Headquarters (GCHQ) and the US National Security Agency (NSA), which work closely together on intelligence and security issues.
“For the US, a major AI powerhouse, to sign this agreement with us, the UK, shows how we are leading the way on AI security,” said Donelan.
He added that with many of the most advanced AI companies now based in the US, the expertise of the American government is key to both understanding the risks of AI and holding companies to their commitments.
However, Donelan insisted that despite conducting research into AI security and ensuring guardrails are in place, the UK has no plans to regulate the technology more broadly in the near future because it is developing so rapidly.
Raimondo argued that AI is “the defining technology of our generation” and said: “This partnership will accelerate the work of both our institutes across the full spectrum of risks, whether to our national security or to our wider society.
Government-industry fusion on AI in the UK
Raimondo argued that by working together, the institutes will better understand AI systems, make more robust assessments and publish more rigorous guidelines.
The UK government-backed AISI, chaired by technology investor and entrepreneur Ian Hogarth, has recruited researchers including Google DeepMind’s Geoffrey Irving and Oxford University’s Chris Summerfield to begin testing existing and unpublished AI models.
OpenAI, Google DeepMind, Microsoft and Meta are among the technology groups that have signed voluntary commitments to open up their latest generative AI models for study by the AISI.
The institute is key to Prime Minister Rishi Sunak’s aim for the UK to play a central role in the development of AI.
The tests will focus on the risks associated with the misuse of the technology, including cybersecurity, drawing on the expertise of the National Cyber Security Centre at GCHQ, according to a person with direct knowledge of the matter.
