Editor’s Note: Concerns over AI being used by the Military Industrial Complex and Intelligence Community to surveil American citizens are legitimate. Before our fellow America First patriots say it’s okay because the administration won’t use it for those purposes, ask yourself two questions: would you object if a Democrat regime were in the White House wielding such power and do you think everyone in the Trump administration is beyond reproach? Then, read this article…
The most consequential corporate moment in the history of artificial intelligence didn’t happen in a boardroom or a congressional hearing. It happened in an all-hands meeting on a Tuesday afternoon, when OpenAI CEO Sam Altman told his own employees — many of them deeply uncomfortable with everything that had unfolded over the previous four days — that their opinions about military strikes do not matter, nor are their concerns over AI being used to surveil American citizens.
“So maybe you think the Iran strike was good and the Venezuela invasion was bad,” Altman told staff, according to a partial transcript of the meeting. “You don’t get to weigh in on that.”
That statement, delivered with the casual confidence of a man who has already signed the contract, marks a turning point in American history that most people haven’t fully processed yet. The nation’s most powerful AI company has handed its technology to the War Department for use in classified operations, acknowledged it will have no say in how that technology is applied in life-and-death situations, and simultaneously told its employees — many of whom helped build that technology with the explicit belief it would not be weaponized without ethical guardrails — to sit down and accept it.
The sequence of events that brought us here moved at a speed that made careful public deliberation nearly impossible. On February 27, the War Department gave Anthropic — the maker of the Claude AI model — a 5:01 p.m. deadline to agree that its technology could be used for “all lawful purposes” without restriction. Anthropic refused, standing firm on two conditions it considered non-negotiable: no use of its AI for fully autonomous weapons systems, and no mass domestic surveillance of American citizens. Secretary of War Pete Hegseth responded by designating Anthropic a “Supply-Chain Risk to National Security” — a label typically reserved for foreign adversaries like Chinese tech firms. President Trump followed with a Truth Social post ordering every federal agency to immediately cease all use of Anthropic’s technology.
Within hours of Anthropic’s blacklisting, Altman announced that OpenAI had reached its own deal with the Pentagon. The timing could not have been more stark. As one company was being punished for holding its line, another rushed into the vacuum. Altman himself would later admit that the move “looked opportunistic and sloppy” and that it was “a good learning experience” for him as OpenAI faces “higher-stakes decisions in the future.”
That admission, buried in a memo to staff he subsequently posted on X, is notable for what it concedes: that the CEO of the world’s most prominent AI company prioritized speed over principle, signed a deal that his own people found ethically troubling, and acknowledged the optics only after the blowback became commercially damaging.
The commercial damage was real. Within 24 hours of the Pentagon deal being announced, U.S. uninstalls of the ChatGPT mobile app jumped 295% compared to the prior 30-day average, according to market intelligence firm Sensor Tower. Downloads of Anthropic’s Claude app rose 37% in the same window. Claude climbed from the 131st most-downloaded app to the top position in the App Store. Consumers, it turned out, had opinions about whether the company behind their AI assistant should be capable of government-sanctioned domestic spying.
The all-hands meeting on Tuesday was Altman’s attempt at damage control, and it raised more questions than it answered. On one hand, Altman told employees that the Pentagon respects OpenAI’s technical expertise, will allow the company to build its own “safety stack” to prevent misuse, and agreed that if a model refuses to perform a task, the government would not force OpenAI to override that refusal. On the other hand, Altman made the power dynamic abundantly clear: operational decisions rest with Secretary Hegseth. OpenAI gets to advise on where its models are a good fit. It does not get to decide what fits.
The distinction matters enormously when you understand what is actually at stake. Claude — Anthropic’s model, not OpenAI’s — was already deeply embedded in the Pentagon’s most sensitive work before all of this began. According to reporting by the Wall Street Journal, Claude was used through Palantir’s platform during the operation to capture former Venezuelan President Nicolás Maduro, and was potentially in use or on standby as U.S. and Israeli forces began airstrikes against Iran. These are not theoretical applications. AI has already touched real military operations with real consequences for real human beings. The question of who controls that AI, and under what constraints, is not an abstract ethics debate.
Anthropic’s position deserves far more credit than it has received in coverage that has largely framed this as a story about corporate stubbornness or AI safety idealism. Anthropic CEO Dario Amodei made clear that his company never raised objections to specific military operations or tried to second-guess individual battlefield decisions. Its objections were structural and narrow: it did not want its models to power systems that kill people without a human making the final decision, and it did not want its models used to conduct mass surveillance of American citizens. Those are not radical positions. They are, in fact, baseline ethical commitments that most Americans would likely endorse if the question were put plainly to them.
The Pentagon’s response to those commitments revealed something important about how the administration views the relationship between government and private industry. Defense officials argued that whether to conduct mass surveillance or deploy autonomous weapons is a legal question — the Pentagon’s legal question, not Anthropic’s. Their position was that once you sell a tool to the military, the military decides how to use it.
Emil Michael, the Pentagon official leading negotiations, called Amodei a “liar” with a “God complex” who was “ok putting our nation’s safety at risk.” Secretary Hegseth declared that “America’s warfighters will never be held hostage by the ideological whims of Big Tech.” These are the words of officials who do not see AI companies as partners in national security — they see them as vendors who should fulfill orders and stay quiet.
That framing has serious implications, and not just for AI companies. The supply chain risk designation that was applied to Anthropic is a legal mechanism typically used to protect the military from adversarial foreign technology. Its application to a domestic American company that refused to remove ethical guardrails has alarmed legal scholars.
Analysis by Lawfare suggests the designation likely “won’t survive first contact with the legal system,” given procedural deficiencies and the fact that Hegseth’s own public statements — calling Anthropic’s position incompatible with “American principles” before any legal process concluded — may constitute evidence of pretext. Anthropic has pledged to challenge the designation in court, calling it “legally unsound” and warning it sets a “dangerous precedent for any American company that negotiates with the government.”
The irony embedded in this entire episode is almost too rich to process. OpenAI, which rushed to fill the void left by Anthropic’s blacklisting, publicly claimed it had secured the same red lines Anthropic was punished for demanding. Altman wrote in his announcement post that prohibitions on domestic mass surveillance and human responsibility for the use of force “are our most important safety principles,” and that the Pentagon “agrees with these principles, reflects them in law and policy, and we put them into our agreement.”
If that is true, then Anthropic was blacklisted not for the substance of its position but for the mere act of insisting that position be written into enforceable contract language. The question of why the Pentagon agreed to accommodate OpenAI but not Anthropic — despite nearly identical stated principles — has not been satisfactorily answered.
One possible explanation involves the competitive dynamics between OpenAI and Anthropic and the pre-existing relationship each company had with the administration. Trump, notably, has had a visible relationship with Altman — who appeared alongside him at the January 2025 Stargate AI infrastructure announcement. Anthropic, by contrast, was founded by former OpenAI employees including Amodei, who has not cultivated the same proximity to this White House. Whether favoritism played a role is speculation, but the appearance is undeniable and the question deserves serious scrutiny.
What is not speculation is the position OpenAI employees now find themselves in. Many of them work on AI safety because they genuinely believe that building powerful AI without robust human oversight is one of the most dangerous things humanity has ever attempted. They watched their CEO sign a classified military contract four days before telling them in a company-wide meeting that their views on military strikes are irrelevant. They watched a competitor get federally blacklisted for holding the same red lines their own CEO publicly endorsed.
Some 70 OpenAI employees had signed an open letter called “We Will Not Be Divided,” expressing solidarity with Anthropic’s position, before their CEO reached around them and signed the deal anyway. That is a profound institutional rupture, regardless of how well Altman handles the follow-up memos.
The broader question this episode forces onto the table is one American society is not remotely ready to answer: who decides how AI is used in warfare? The Pentagon’s position is clear — the military decides. AI companies, in their view, are utilities. The contractors who supply electricity to military bases do not get veto power over how the electricity is used.
But AI is not electricity. AI systems make decisions, surface targets, process surveillance data, and in increasingly autonomous configurations, may do so faster than any human can review. The ethical load embedded in those systems does not disappear simply because the government purchased access to them. It transfers to whoever wrote the values — or the absence of values — into the model’s architecture.
Sam Altman ended his all-hands by telling employees something that sounds reassuring in isolation: “Things are moving so fast that we need to urgently educate the world so that the democratic process has time to catch up.”
But here is the problem with that framing. The democratic process did not catch up before OpenAI signed a classified military contract. It did not catch up before Claude was reportedly used in operations targeting Venezuela and Iran. And it will not catch up before the next operation, or the one after that.
The technology has already outpaced the governance. Altman’s admission of that gap is honest, and to his credit he is saying it out loud. But no one in that meeting — not the employees, not the CEO — gets to slow down the clock. The military has the contract. Operational decisions belong to Hegseth. And the AI is already deployed.
