Anthropic held talks with the Trump administration to resolve a dispute with the Pentagon and grant access to a new Mythos AI model.
Axios reported on this.
According to sources, Dario Amodei, the company’s chief executive, was to hold a meeting in the West Wing of the White House with White House Chief of Staff Susie Wiles.
One source close to the negotiations said it would be “extremely irresponsible” to deprive the U.S. government of access to this model’s technological capabilities, as it would be a “gift to China.”
The Donald Trump administration recognizes the potential of Anthropic’s new Claude model, called Mythos, which has advanced cyber capabilities, including the ability to overcome digital defense mechanisms.
At the same time, some parts of the U.S. intelligence community, as well as the Cybersecurity and Infrastructure Security Agency, are already testing Mythos. The U.S. Treasury Department has also shown interest in the model.
Mythos
Mythos is an experimental next-generation AI model from Anthropic, announced on April 7, 2026. Its key feature is a very high level of cybersecurity capability. At the same time, the model’s ability to analyze code security, according to the company, is not the result of specialized training but derives from strong general programming skills and logical reasoning.
The company decided not to release Mythos to the public because of its advanced capabilities for cyberattacks, which could make it a powerful tool for hackers, criminal groups, and hostile states.
As part of the Glasswing project, the company provides limited access through government channels and trusted partner companies to search for vulnerabilities. According to Anthropic, within a few weeks in offline mode, the model discovered thousands of previously unknown vulnerabilities in every major operating system and browser.
Dispute between Anthropic and the Pentagon
In February 2026, it was announced that the U.S. military had successfully used Anthropic artificial intelligence during a secret operation to detain Venezuelan President Nicolas Maduro.
This mission was the first confirmed use of Anthropic models in covert operations by the U.S. Department of War. This incident provoked a sharp conflict between the developer and the U.S. government. Anthropic’s rules officially prohibit the use of Claude to promote violence, develop weapons, or conduct surveillance.
Shortly afterward, Dario Amodei refused to lift restrictions on the use of Anthropic technologies, despite threats from Pentagon chief Pete Hegseth to stop cooperation with the company. In response, Anthropic filed a lawsuit against the U.S. Department of War after the company was blacklisted.
