Anthropic model flags vulnerabilities in classified U.S. systems during government tests

Traders Union
All News
Financial News
Anthropic AI flags U.S. vulnerabilities

Highlights

Anthropic’s Mythos model, tested through Project Glasswing, identified vulnerabilities in classified U.S. systems within hours during a government-led security exercise.
Democratic Sen. Mark Warner stated that Mythos broke into almost all classified systems in hours, citing Gen. Joshua Rudd of the NSA and U.S. Cyber Command.
The Trump administration ordered Anthropic to block foreign nationals from Fable 5 and Mythos 5 access and temporarily disabled the models, sparking industry backlash over national security implications.

Government testing and security findings

As first reported by The Associated Press, a U.S. official says Anthropic worked with U.S. intelligence agencies in a testing exercise that used the company’s Mythos model to examine highly secure government systems. The official, who spoke anonymously, says the model identified certain vulnerabilities within hours, though that did not mean it exploited them in that time.

The testing was conducted through Anthropic’s Project Glasswing initiative, which brings together large technology companies and other groups to help protect critical software. The effort is presented as a response to what the company sees as potentially severe risks that the Mythos model could pose to public safety, national security and the wider economy.

Democratic Sen. Mark Warner of Virginia briefly referred to the exercise during a June 11 hearing before the Senate Committee on Banking, Housing, and Urban Affairs. Warner says the tool broke into almost all classified systems in hours rather than weeks, attributing that information to Gen. Joshua Rudd, who leads the National Security Agency and U.S. Cyber Command.

The NSA declines to comment in an email, and an Anthropic spokesperson also declines to comment.

Policy tensions around Anthropic’s AI models

Even as Anthropic cooperates with U.S. agencies on security testing, tensions are growing between the California company and the Trump administration. Anthropic has raised concerns about how the U.S. military would use its AI, while the administration has moved to restrict access to some of the company’s models.

Earlier this month, the administration issued a directive requiring Anthropic to prevent foreign nationals from using its latest AI models, Fable 5 and Mythos 5. Anthropic released Fable widely earlier this month, while access to the more advanced Mythos remains tightly limited because of cybersecurity concerns.

The directive follows an executive order signed by President Donald Trump 10 days earlier to create a framework for the federal government to review national security risks from the most advanced AI systems for up to a month before public release. The order says participation by AI developers is voluntary.

Anthropic says it disabled the models for all customers to comply with the directive, while also saying the government’s response was not warranted by the potential security issue it had raised. A group of cybersecurity executives has also asked the administration to withdraw the directive, arguing it could help U.S. adversaries more than it hurts them.

In a letter to the government, more than 100 cybersecurity experts and leaders from companies including Adobe and Nvidia say Anthropic’s Mythos models are quite good at finding software flaws and weaponizing exploits, but not uniquely good at those tasks.

Our earlier report on Legion LegalTech’s federal lawsuit over the June 12 AI access restriction explained how a U.S. trade directive required Anthropic to disable Fable 5 and Mythos 5 for foreign nationals, prompting the company to shut off access for all customers. Legion said the cutoff immediately disrupted its Canada-based development team and threatened its operations, highlighting how export-style limits on frontier AI can quickly ripple through businesses that rely on a small number of model providers.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.