White House reverses course and restores UK AI Safety Institute's access to frontier models
decoder ·
The White House has reversed its previous decision to block the UK's AI Safety Institute (AISI) from accessing frontier artificial intelligence models. This restoration of access allows the institute to continue testing systems for safety risks prior to their commercial release. The reversal follows diplomatic pressure exerted by UK AI Minister Kanishka Narayan after the Trump administration had initially instructed companies like Anthropic, Google DeepMind, and OpenAI to cease sharing model data with AISI. Consequently, Anthropic had previously restricted access to its new Claude Mythos 5.1 model during the blockade period. In response to this development, Minister Narayan plans to announce new AI safety measures, including stricter reporting requirements for security incidents involving critical infrastructure. This policy shift occurs amidst broader tensions regarding international AI regulation, highlighted by Prime Minister Andy Burnham's push for UN involvement and President Trump's warnings against globalist organizations interfering in US development. Additionally, UK intelligence heads have recently warned that AI is increasingly empowering hostile actors and terrorists.
decoder