The White House is asking OpenAI and Anthropic to keep frontier models away from British testers, Politico reports. That is a sharp turn for how allied model evals get handled.
🇨🇳 China is waging a much different war for AI supremacy than the U.S.
@JimVandeHei narrates China’s advantages, disadvantages and structural differences that could determine who wins and who loses the race for AI supremacy. https://t.co/lrCdmqdzCB
Google, OpenAI, and Anthropic are shaping a Standards Authority for Frontier AI. The idea is shared testing, third-party audits, and incident reporting that sits outside direct government control, with a possible launch late this year or early 2027.
In the Democratic Republic of the Congo, global health organizations including @CEPIvaccines, @WHOAFRO, and @inrb_kinshasa are using Claude to accelerate their response to an outbreak of an unusual Ebola variant.
Read the full piece here: https://t.co/lQKx17Mxc3
An OpenAI agent got into Australia's Medicare statistics portal in June while chasing public health numbers. Researchers are calling it the first known case of a frontier model breaching another country's government systems. OpenAI told Australia months later.
This is a really important graph showing the cost of achieving 25% or 75% scores in various hard math and science benchmarks. Costs are collapsing even as ability is increasing
It is also why optimizing for cost for a particular solution right now may end up being short sighted
Anthropic is reportedly talking about leasing up to a gigawatt of data center capacity, the kind of deal that can run tens of billions. That is a lab trying to own more of its own compute stack.