At least one current Anthropic shareholder has raised pointed concerns about CEO Dario Amodei's temperament and public conduct, worries that carry real weight at a company now valued at roughly $380 billion with Amazon and Google among its backers. The complaint, as the New York Post reported, centers on a claim that the 43-year-old cofounder "cannot control his emotions", a damaging charge for any executive steering one of the fastest-growing companies in history through a high-stakes dispute with the Pentagon.
The friction burst into public view last month, when an internal memo Amodei sent to employees after the Pentagon blacklisted Anthropic was leaked. In it, the CEO took aim at both President Trump and rival Sam Altman of OpenAI, language that startled even some allies inside the company.
For conservative readers who believe the defense establishment should work with the best available technology, and that CEOs of major government contractors should exercise discipline rather than air personal grudges, the Anthropic saga is a case study in what happens when Silicon Valley ideology collides with national security reality.
The Pentagon's decision to blacklist Anthropic came after the company refused to remove safeguards on how its Claude AI chatbot can be used in military settings. A top Pentagon official defended the move, stating Claude was trained "using a fundamentally different ideology from what the Pentagon wants for its systems."
Hours after that blacklist decision, Amodei fired off an internal message to staff. He accused OpenAI's Altman of offering "dictator-style praise to Trump" and dismissed unnamed parties' claims as "straight up lies." He called certain people a "gullible bunch."
The memo leaked. Amodei later apologized and insisted that "Anthropic did not leak this post nor direct anyone else to do so." He added that it was "not in our interest to escalate this situation."
But the damage was done. The unnamed shareholder's concern, that Amodei cannot govern his own reactions, gained traction precisely because the memo confirmed the pattern rather than contradicting it. When your company is suing the Pentagon and your investors include two of the largest corporations on Earth, a leaked screed attacking a sitting president and your chief competitor is not a minor embarrassment. It is a governance problem.
The Amodei-Altman feud did not start with the Pentagon. It traces back to OpenAI itself. Dario Amodei and his sister Daniela, a fellow cofounder of Anthropic, left OpenAI in 2021 after disagreements with Altman over the direction of AI safety research. The Wall Street Journal last week reported on workplace tensions involving the Amodei siblings, Altman, and OpenAI cofounder Greg Brockman.
One flashpoint reportedly came in late 2020, when Altman accused the Amodei siblings of urging colleagues to badmouth him to OpenAI's board. Concerns within the group "went into overdrive" in 2022, after Altman decided to release ChatGPT to the public, a move the safety-focused Amodei camp viewed with alarm.
The personal dimension of this rivalry was on display at a February tech conference in India, where Amodei and Altman were placed next to each other on stage alongside Indian Prime Minister Narendra Modi and Google CEO Sundar Pichai. The two refused to link hands for a group photo. That kind of pettiness might be forgiven in a startup founder. In a CEO whose company holds defense contracts and counts sovereign-scale investors, it reads differently.
The broader upheaval AI is causing across markets and policy makes these personal vendettas more than gossip. When the technology you build could reshape military operations, the adults in the room need to act like it.
Anthropic's clash with the Defense Department is the most consequential test Amodei has faced. The company has branded itself as the "safety-first" AI lab, a positioning that won it admirers in Washington's regulatory class but also drew sharp criticism from figures like former White House AI czar David Sacks, who has alleged Anthropic's safety branding amounts to "regulatory capture."
The Pentagon designated Anthropic a "supply chain risk" after the company declined to strip guardrails from Claude for military applications. On the eve of a Pentagon deadline, Amodei publicly announced that Anthropic could not "in good conscience accede to their request." The company then sued the Pentagon, calling the government's actions "unprecedented and unlawful."
A federal judge issued a preliminary injunction blocking the government ban on March 26, giving Anthropic a legal lifeline. But the broader question, whether a company that refuses to adapt its product for its largest potential customer can sustain a $380 billion valuation, remains wide open.
The Pentagon contract dispute exposes a tension at the heart of Anthropic's business model. You can position yourself as the responsible AI company, or you can be a major defense contractor. Doing both requires a CEO with the discipline to navigate competing pressures without alienating either side. The leaked memo suggests Amodei may lack that discipline at the moment it matters most.
Inside Anthropic, two figures have historically served as counterweights to Amodei's impulses: his sister Daniela and policy chief Jack Clark. Unnamed sources told the Post that the pair usually tried to restrain the CEO's instincts. Whether that restraint held during the Pentagon crisis, or whether Amodei overrode it, is unclear.
Amodei also hosts internal meetings called "Dario Vision Quests," or DVQs. The name alone suggests a corporate culture built heavily around one personality. That can work when the founder's instincts are sound. When they are not, there is no institutional buffer.
The pattern is familiar across Silicon Valley, where investor anxiety over AI leadership has already cost software executives tens of billions in market value. Founder-worship works on the upswing. On the downswing, it leaves shareholders exposed to exactly the kind of unforced errors Amodei's memo represents.
Several Anthropic backers, including Altimeter, Menlo Ventures, and Spark Capital, publicly supported the company following its dispute with the Pentagon. That show of solidarity may reflect genuine conviction, or it may reflect the fact that investors in a $380 billion private company have limited options once the money is in.
The unnamed shareholder's complaint stands out precisely because it breaks ranks. Telling a reporter that your CEO "cannot control his emotions" is not something investors do casually. It signals that at least some of the money behind Anthropic sees a leadership risk that public statements of support cannot paper over.
Amodei once warned publicly that AI could drive unemployment to 20 percent. Whether or not that forecast proves accurate, it reflects the scale of consequences attached to the decisions his company makes. Investors are right to ask whether the person making those decisions can keep his composure when the pressure is real.
The stakes extend well beyond one company. As Wall Street grapples with mounting doubts about whether AI investments will pay off, the conduct of the industry's most prominent leaders matters more, not less. A CEO who picks public fights with the Pentagon, leaks bitter memos about rivals, and refuses to shake hands at international conferences is not projecting the kind of stability that justifies a $380 billion price tag.
Anthropic's refusal to modify Claude for military use deserves scrutiny on its own terms. The company frames it as a principled stand for AI safety. The Pentagon frames it as ideological stubbornness that undermines national defense. David Sacks frames it as a play for regulatory advantage, "regulatory capture" dressed up as ethics.
All three frames may contain some truth. But the leaked memo complicates Anthropic's claim to the moral high ground. A CEO who privately attacks the president and his chief competitor in intemperate language, then apologizes when the memo goes public, is not demonstrating the measured judgment that "safety-first" branding implies.
The question is not whether AI companies should have safety standards. They should. The question is whether Anthropic's particular version of safety, one that apparently prevents the company from serving the United States military, reflects genuine technical caution or a political disposition that happens to align with the preferences of its founder.
Even some in the broader tech world have noticed how AI pressures are reshaping executive leadership across industries. The difference is that most CEOs facing those pressures do not respond by firing off internal screeds about their rivals.
The preliminary injunction gives Anthropic breathing room, but the lawsuit against the Pentagon is far from resolved. The company's valuation, its relationships with Amazon and Google, and its ability to compete with OpenAI all depend on whether Amodei can project stability going forward.
Open questions remain. What specific safeguards did the Pentagon want removed? What court issued the injunction, and on what grounds? How many other shareholders share the unnamed investor's concerns? None of these answers are publicly available yet.
What is available is a pattern: a CEO who built a company on the promise of responsible AI, then responded to the biggest challenge of his tenure with a leaked memo full of personal attacks, a public refusal to cooperate with the Defense Department, and a lawsuit against the government, all while his investors watched.
Billions of dollars and the future of American AI defense capability are riding on one man's ability to keep his composure. So far, the record is not reassuring.