White House asks OpenAI and Anthropic to let US test new models before UK
The US government asked OpenAI and Anthropic not to give new frontier models to the UK's AI Security Institute until they have been tested by the US government, and Anthropic appears to have complied.
🇺🇸🇬🇧 The White House is making a rather consequential demand of America’s biggest AI labs: show us the new models before you show Britain. The administration asked OpenAI and Anthropic not to give new frontier models to the U.K.’s AI Security Institute until they have first been tested by the U.S. government. Anthropic appears to have complied. Its new Claude Mythos 5.1 was initially restricted to U.S. organizations, while OpenAI had previously given Britain prerelease access to GPT-6 Astra. The official explanation is cybersecurity. Washington wants time to understand what increasingly powerful models can do, and protect U.S. systems, before handing them to allies. But there is something larger happening here. The most advanced AI models are starting to be treated less like ordinary software releases and more like strategic technology. Britain wants global rules. But Washington wants to get the first look. Source: POLITICO / Writer: Oliver