Report: China seeking access to Anthropic’s Mythos model
Anthropic’s unreleased AI model Mythos has sent shock waves through companies and governments around the world, fearful of what the model will mean for cybersecurity. Even the US Treasury department scrambled to secure access to harden its defenses ahead of a wide release.
Anthropic is currently only sharing access to Mythos to a short list of companies and government agencies.
The New York Times is reporting that China is seeking access to Mythos as well, setting off alarms in the White House. At a Singapore conference last month, an employee from a Chinese think tank reportedly approached representatives from Anthropic, seeking access to Mythos — a move that was interpreted in Washington as a potential effort to secure access for the Chinese government. According to the report, Anthropic declined that request.
As AI models rapidly gain powerful new capabilities, the US government is wrestling over what kinds of controls (if any) it should apply to prevent American technology from being used by our rivals.
The Washington Post is reporting that an executive order from the Trump administration may be imminent that would allow US intelligence agencies to evaluate new AI models before release.
The New York Times is reporting that China is seeking access to Mythos as well, setting off alarms in the White House. At a Singapore conference last month, an employee from a Chinese think tank reportedly approached representatives from Anthropic, seeking access to Mythos — a move that was interpreted in Washington as a potential effort to secure access for the Chinese government. According to the report, Anthropic declined that request.
As AI models rapidly gain powerful new capabilities, the US government is wrestling over what kinds of controls (if any) it should apply to prevent American technology from being used by our rivals.
The Washington Post is reporting that an executive order from the Trump administration may be imminent that would allow US intelligence agencies to evaluate new AI models before release.
