The US Government Is Suing Anthropic and Begging to Use Its AI at the Same Time
The US Government Is Suing Anthropic and Begging to Use Its AI at the Same Time
Anthropic co-founder Jack Clark walked onto the stage at the Semafor World Economy Summit on April 14, 2026, and confirmed something that sounds like a typo: his company is actively briefing the Trump administration about its new Mythos model, even though the Pentagon has labeled Anthropic a supply-chain risk and cut off all business ties.
Not a metaphor. Both things are happening at once.
Mythos, announced April 7, is Anthropic’s most capable model for coding and agentic tasks. That’s the company’s own description. The model is not available to the public. Anthropic says it’s limiting access because Mythos, despite not being trained for cybersecurity work, happens to be dangerously good at finding software vulnerabilities. Whether that’s genuine caution or a smart enterprise sales strategy depends on who you ask.
What Actually Happened
Before Mythos launched, Vice President JD Vance and Treasury Secretary Scott Bessent summoned tech CEOs to discuss AI security and cyber threats. That meeting, reported by CNBC on April 10, set the stage for what came next.
After the launch, Bessent and Federal Reserve Chair Jerome Powell called bank executives in for another session. This time the message was different. According to Bloomberg, officials encouraged those banks to test Mythos for vulnerability detection. JPMorgan Chase was already listed as an early partner. Goldman Sachs, Citigroup, Bank of America, and Morgan Stanley are now reportedly testing it too.
Five of the largest banks in the United States, quietly urged by regulators to adopt a model from a company the government is simultaneously fighting in court.
The Court Fight
Anthropic sued the Department of Defense in March 2026 after the Pentagon designated the company a supply-chain risk. The designation followed failed negotiations over military access to Anthropic’s AI systems. Anthropic wanted limits on how its models could be used, specifically around mass surveillance of Americans and fully autonomous weapons. The Pentagon wanted unrestricted access. OpenAI ended up winning that contract instead.
A San Francisco federal judge later granted Anthropic a preliminary injunction blocking the Trump administration from enforcing a ban on Claude. But on April 8, an appeals court dealt Anthropic a loss, refusing to temporarily block the Pentagon’s blacklisting.
Clark downplayed the dispute at the Semafor event, calling it a “narrow contracting dispute” and saying Anthropic didn’t want it to interfere with national security engagement. His exact words: “Our position is the government has to know about this stuff, and we have to find new ways for the government to partner with a private sector that is making things that are truly revolutionizing the economy.”
Why UK Regulators Are Paying Attention
The Financial Times reported that UK financial regulators are also discussing risks posed by Mythos. The model’s ability to find zero-day vulnerabilities at scale has governments on edge, and not just in Washington. A tool that can audit code for exploits faster than any human team is either the best defensive weapon a bank could ask for or a serious offensive capability in the wrong hands.
Washington Post columnist wrote on April 12 that the Pentagon’s ban on Anthropic looks “shortsighted” given Mythos’s ability to detect critical software flaws. The headline practically wrote itself.
The Employment Question Nobody Can Answer
Clark also addressed AI’s impact on jobs at the summit. Anthropic CEO Dario Amodei has previously warned that AI advances could push unemployment to Depression-era levels. Clark, who leads a team of economists at Anthropic, was more measured. He said the company is currently seeing “some potential weakness in early graduate employment” across select industries. Not mass displacement yet, but early tremors.
That’s worth watching. When the company building the technology admits entry-level white-collar jobs are starting to soften, it’s probably not marketing talk.
What Makes This Strange
The whole situation is a contradiction that tells you something about where AI policy actually stands in 2026. The Department of Defense says Anthropic is too risky to do business with. The Treasury Department and the Fed are nudging banks to adopt Anthropic’s product. The appeals court says the Pentagon can blacklist Anthropic. A lower court says the government can’t ban Claude. The company is suing the government and briefing it on frontier models at the same time.
There is no coherent policy here. There are competing fiefdoms inside the US government pulling in opposite directions, and Anthropic is caught in the middle, playing every angle because that’s what a company worth tens of billions of dollars does when the rules are this muddled.
Clark confirmed Anthropic will brief the government on future models too. The relationship is broken but the conversations haven’t stopped. That’s probably the most honest description of US AI governance right now: broken, but still talking.
For tool-by-tool comparisons, see our AI coding listings and the comparisons section.