Key Points:
- Anthropic co-founder Jack Clark confirmed the company is discussing its new Mythos AI model with the Trump administration.
- The Pentagon recently labeled Anthropic a supply-chain risk and barred contractors from using its software.
- The dispute started when Anthropic demanded strict guardrails on how the military could use its artificial intelligence tools.
- Cybersecurity experts warn that the new Mythos model possesses an unprecedented ability to identify and exploit computer vulnerabilities.
Anthropic is actively discussing its newest artificial intelligence model with the Trump administration. The technology company chose to keep the lines of communication open despite a massive, ongoing legal fight with the government. Last month, the Pentagon officially cut off all business with the US-based AI firm. The military completely banned the software after a bitter dispute over how the armed forces could use the technology.
The massive falling-out between the tech company and the military began over safety rules. Anthropic demanded strict guardrails limiting exactly how the military could deploy its artificial intelligence tools in the field. The Pentagon strongly rejected these restrictions. Because the two sides could not reach an agreement, the military agency officially designated Anthropic a supply chain risk. This harsh label legally barred the Pentagon and all of its outside contractors from using any Anthropic software.
Despite the ban, Anthropic wants to repair its relationship with Washington. Jack Clark, the co-founder of the AI company, spoke at the recent Semafor World Economy event. He downplayed the fight, calling it a narrow contracting dispute. Clark stressed that he does not want this argument to hide the fact that Anthropic cares deeply about national security. He firmly believes the US government needs to know about the company’s latest technological breakthroughs.
Clark confirmed the ongoing discussions during his public appearance. He stated clearly that his team is absolutely talking to the government about their new Mythos model right now. He also promised that the company will continue to brief federal officials on all future models it develops. However, the exact details of these meetings remain a secret. Clark did not reveal which specific government agencies are participating in the talks.
The new software model is incredibly powerful. Anthropic officially announced the Mythos model on April 7 through a company blog post. The tech firm described Mythos as its most capable model yet for complex coding and agentic tasks. Agentic tasks refer to the software’s ability to act autonomously, making decisions and solving problems without constant human guidance.
This autonomous power scares some cybersecurity experts. Because Mythos can write computer code at such a high level, it possesses an unprecedented ability to hunt for weaknesses in existing computer systems. Experts warn that the software could easily identify critical cybersecurity vulnerabilities across the internet. Even worse, the AI could quickly devise creative new ways to exploit those exact weaknesses if it fell into the wrong hands.
Because of these massive security risks, the legal fight between Anthropic and the Pentagon continues in federal court. Just last week, a federal appeals court in Washington, D.C., ruled on the matter. The judges declined to block the Pentagon’s decision to blacklist Anthropic. This ruling was a major legal victory for the Trump administration, allowing the military to keep the ban in effect for now.
The legal situation remains highly complicated and confusing. The recent victory for the Trump administration came just days after a different appeals court reached the exact opposite conclusion. That separate court ruled in favor of Anthropic during a related legal challenge. Until the courts finally agree on a single legal standard, Anthropic will likely remain blacklisted by the military while simultaneously briefing the Trump administration on its most powerful technology.
