“We are in a distinctly unique place globally as the only country, outside of the countries where the labs and models are, which gets access pre-deployment in the wild,” said Narayan.

The U.K.’s AISI “not just keeps the British public safe by virtue of testing, but also shares insights with the Americans so that the American public is kept safer,” said Narayan. It puts the U.K. in a “very unique position, and I think that position has not diminished in the slightest.”

Last week U.K. prime minister Andy Burnham used his address at UNGA to announce that the U.K. would use its presidency of the G20 next year to broker global standards on AI.

Regarding the U.K.’s global standing, Narayan said “Who do you trust to have both capability and the honesty to be a broker on a broader global conversation on this?” He said there are “very few” countries that would make such a list, “but I can bet you that the United Kingdom is pretty much top of that list.”

Narayan also warned that the U.K. has “imported American politics” around data centers. In the U.S. political backlash to data centers has been intense, but Narayan said the U.K. buildout has been more considered and not on the same scale.

He addressed calls from lawmakers to ban development of so-called ‘superintelligence,’ a hypothetical future form of AI with more advanced thinking skills than any human. Narayan said that superintelligence is already impossible to build in the U.K., as the country’s copyright laws mean it’s already "illegal" to build cutting-edge AI large language models.

“What further law would you need, given that it is currently illegal to develop a frontier large language model ... given the copyright position that we have in this country and, frankly, given the overall compute situation that we have in this country?”