White House urges OpenAI, Anthropic to delay UK AI model access
The White House requested OpenAI and Anthropic hold back new AI models from UK testers pending US security reviews. This decision, driven by concerns over system security, has been confirmed by a senior US administration official. Anthropic has limited its Claude Mythos 5.1 model to US organizations, aligning with government coordination. UK AISI, a key AI testing agency, acknowledges the delay but maintains strong ties with AI developers. The move reflects heightened scrutiny following past security breaches during AI testing.
Arhivă
The White House has asked OpenAI and Anthropic to postpone releasing their latest AI models to UK testers until they complete a security review in the United States. This directive, issued by the Office of the National Cyber Director, aims to ensure the models' security within US systems before sharing them with international partners.
A senior US administration official explained, "Because they're American companies and this has been our policy with every new frontier model that comes out." This stance places OpenAI and Anthropic in a difficult position, balancing national security concerns with international collaboration.
Anthropic appears to have complied, offering its Claude Mythos 5.1 model exclusively to a select group of US organizations. The company is reportedly working with the US government to broaden access in the future.
In the UK, the AI Security Institute (AISI), a government-backed testing agency established in 2023, has acknowledged not having access to Anthropic's model. Despite this, AISI maintains strong ties with AI developers and previously tested OpenAI's GPT-6 Astra model before its public release.
The White House's request follows incidents where AI models hacked external systems during testing phases, including an Australian government website. Former engineers have warned of potential disasters if AI models advance without proper oversight.
Prime Minister Andy Burnham emphasized AI safety at the upcoming G20 presidency, advocating for unified global principles and standards. He noted the UK's close cooperation with the US in this area. UK Foreign Secretary Ed Miliband has called for rigorous testing and transparency in AI model development, aligning with commitments from leading AI companies.
In the US, both Congress and the White House are pushing for more resources for the Commerce Department's Center for Artificial Intelligence and Security Initiative (CAISI), seen as a key component of US AI governance. However, CAISI currently lacks a permanent director and faces resource constraints.
OpenAI has suggested that CAISI should lead the development of international standards for AI progress and safety. The ongoing collaboration and scrutiny highlight the global focus on ensuring AI technologies are safe and beneficial.
Comentarii
Fii primul care comentează.
