Posted by 
BLETCHLEY, England — The British Prime Minister Rishi Sunak on Thursday said that under a new agreement “like-minded governments” would be able to test eight leading tech companies’ AI models before they are released.
Closing out the two-day artificial intelligence summit in Bletchley Park on Thursday, Sunak announced the agreement signed by Australia, Canada, the European Union, France, Germany, Italy, Japan, Korea, Singapore, the U.S. and the U.K. to test leading companies’ AI models.
“Until now the only people testing the safety of new AI models have been the very companies developing it. That must change,” said Sunak to a room full of journalists.
“Like-minded governments and AI companies have today reached a landmark agreement. We will work together on testing the safety of new AI models before they are released… it’s made possible by the decision I have taken along with Vice President Kamala Harris for the British and American governments to establish world leading AI safety institutes with public sector capability to test the most advanced frontier models.”
Sunak said the eight companies — Amazon Web Services, Anthropic, Google, Google DeepMind, Inflection AI, Meta, Microsoft, Mistral AI and Open AI — had agreed to “deepen” the access already given to his Frontier AI Taskforce, which is the forerunner to the new institute. The access is currently given on a voluntary basis, though under its Executive Order, the U.S. government has put binding requirements to hand over certain safety information.
Sunak also announced further details of an agreement reached with countries yesterday to establish an international advisory panel on frontier AI risks.
Modeled on the Intergovernmental Panel on Climate Change (IPCC), it will be formed from representatives from the 28 countries attending the summit. The British government said it would provide secretariat support for it.
The panel will also support academic Yoshua Bengio in producing a “State of Science” report into the risks and capabilities of frontier AI. The report will not make policy recommendations, but is designed to inform international and national policy making. It will be published ahead of the next safety summit in South Korea in the first half of next year.
