Anthropic sends top engineers to Washington to resolve Mythos dispute
What's the story
Anthropic, a leading artificial intelligence (AI) company, has sent its top engineers to Washington. The move comes as part of the company's efforts to resolve an ongoing dispute with the Trump administration over restrictions on some of its most advanced AI models. According to Axios, the discussions between Anthropic's engineers and White House officials are aimed at addressing concerns that led to the suspension of their flagship AI models, Claude Mythos 5 and Claude Fable 5.
Order impact
Anthropic had to disable access to AI models
The move to send engineers comes days after Anthropic had to disable access to its two models. This was due to a federal export control order that required the company to block access for non-US nationals. The sudden service shutdown affected paying Claude subscribers who had just started using these advanced AI systems.
Security issues
Why were the models banned?
The restrictions were imposed due to concerns raised by the US government over the capabilities and security of Anthropic's latest AI systems. Mythos 5 and Fable 5 are part of the company's "Mythos-class" family, which is built on the same technology foundation as Claude Mythos Preview. The earlier model was deemed too powerful for public release due to potential misuse risks.
Bypass risks
Concerns over potential 'jailbreaks'
Despite the introduction of additional safeguards with Mythos 5 and Fable 5, concerns were raised shortly after their launch. Reports indicate that several companies, including Amazon, warned White House officials about potential ways to bypass some of these models' protections using techniques like jailbreaks. The issue has quickly escalated into a major flashpoint in the AI industry, raising questions over government regulation of increasingly powerful artificial intelligence systems.
Security warning
National security risks
Reports suggest that White House officials warned Anthropic on Friday about the potential national security risk posed by keeping these models online. The company was given a short window to respond before a formal export control order. Within hours, access to the models was disabled.
Access fears
Potential China-linked access
A Semafor report also revealed that US officials were looking into the possibility of a China-linked group gaining access to a Mythos-class model. However, Anthropic does not officially allow access to its AI systems from China. The Axios report further indicates that the administration was unhappy with Anthropic's initial response to complaints about potential vulnerabilities, which may have contributed to the aggressive action taken against the company.