White House Challenges Anthropic’s Plans for AI Expansion
The White House is reportedly voicing concerns over Anthropic’s intentions to broaden access to its AI tool, Claude Mythos. Executives from the company have warned that, in the wrong hands, this powerful technology could potentially lead to hacking incidents and even terrorist activities.
Recently, Anthropic suggested extending access to Mythos to an additional 70 companies, raising the total to 120, as detailed by sources familiar with the matter.
Earlier in the month, Anthropic introduced Project Glasswing, which aims to provide access to the model for a select group of companies, including major players like Amazon, Google, and JPMorgan.
According to insiders, White House officials have communicated their opposition to this expansion, primarily due to security risks.
Anthropic has publicly analyzed the potential consequences of a breach, indicating that if Mythos were compromised, it could disrupt critical infrastructure such as power grids and hospitals.
There are also concerns among some officials from the Trump administration regarding Anthropic’s capacity to serve both government entities and private businesses effectively.
White House representatives have indicated they are actively collaborating with the private sector to strike a balance between fostering innovation and maintaining national security.
Anthropic has not responded to requests for comments.
Discussions between officials from the White House and Anthropic executives regarding Mythos are perceived as an effort to improve relations. Currently, the two parties are engaged in legal disputes in separate cases.
Earlier this year, the Pentagon ended its contract with Anthropic and threatened to place the company on a blacklist after Anthropic declined to grant unrestricted access to its AI tools, especially regarding mass surveillance and military applications.
On April 17, Dario Amodei, CEO of Anthropic, met with Susie Wiles, the President’s chief of staff, and other high-ranking officials.
Some government bodies have already received access to Mythos, and the Trump administration aims to extend those permissions.
Last week, Anthropic raised alarms about potential unauthorized access to Mythos, prompting fears about the possible fallout if the AI were fully operational.
Reports indicated that a limited number of users had breached Mythos, though Anthropic clarified that the tool was only available to selected enterprise clients.
Further, it was revealed that these unauthorized users have been utilizing Mythos frequently, but not in a cybersecurity context.
This AI technology is adept at identifying and exploiting vulnerabilities, prompting Anthropic to offer early access to some security researchers as a precautionary measure.
In addition, the company has recently entered into agreements with several firms, including Amazon, Google, and Broadcom, aiming to bolster its computational capacity, driven by concerns about restricted computing resources hindering growth.
