An Ultimatum With Consequences for Everyone
On February 24, 2026, Defense Secretary Pete Hegseth delivered an ultimatum to Anthropic CEO Dario Amodei: remove all restrictions on how the Pentagon can use Claude, the company’s AI system, by Friday — or face the termination of a $200 million defense contract and potentially far worse consequences.
The demand is simple. The implications are not.
Anthropic is the only AI company currently approved to operate on classified US military networks. It got there through a partnership with defense contractor Palantir, which integrated Claude into its intelligence platforms for military and intelligence use. Now the Pentagon wants that access without limits — and the company building the AI is saying no.
The standoff has escalated into the most consequential confrontation between a technology company and the US government since Apple refused to build an FBI backdoor into the iPhone in 2016. But the stakes this time are higher, the tools more powerful, and the potential for abuse far greater.
What Anthropic Is Refusing to Do
Anthropic has drawn two red lines:
No autonomous weapons. Anthropic refuses to allow Claude to be used in fully autonomous targeting systems — weapons that select and engage targets without a human making the final decision. Amodei has argued that “the constitutional protections in our military structures depend on humans who would disobey illegal orders — with fully autonomous weapons, that safeguard disappears.”
No mass surveillance of Americans. Anthropic refuses to allow Claude to be used for large-scale domestic surveillance — monitoring the communications, movements, or associations of US citizens without individualized warrants.
These aren’t abstract ethical positions. They’re responses to capabilities that already exist. A system like Claude, with access to classified intelligence databases, could process billions of communications, identify patterns of dissent, flag individuals for investigation, and do it at a scale that no human workforce could match. The question is whether it should.
What the Pentagon Is Threatening
Hegseth’s threats go well beyond canceling a contract. The Pentagon has signaled it may:
Designate Anthropic a “supply chain risk.” This designation — typically reserved for foreign adversaries like Huawei and Kaspersky — would effectively blacklist Anthropic from all government work and force every defense contractor to remove Claude from their systems. It would be an existential threat to Anthropic’s government business.
Invoke the Defense Production Act. The DPA, originally passed during the Korean War, gives the President extraordinary power to compel private companies to produce goods and services for national defense. Invoking it against an AI company would set a precedent that the government can force any technology company to provide its products for military use, regardless of the company’s objections, safety concerns, or ethical principles.
Hegseth has publicly championed AI systems operating “without ideological constraints that limit lawful military applications” — framing safety guardrails as political bias rather than engineering prudence.
The Venezuela Operation: A Preview of What’s Coming
The confrontation didn’t start with an ultimatum. It started with a raid.
On January 3, 2026, US special operations forces captured Venezuelan President Nicolás Maduro in an operation that killed 83 people, including 47 soldiers. The Wall Street Journal reported that Claude was used during the operation through the Palantir integration on classified military networks.
During a subsequent meeting between Palantir and Anthropic staff, an Anthropic employee reportedly expressed concerns about the operation. A Palantir executive worried that Anthropic might disapprove of the raid. Pentagon officials characterized this — an employee expressing concern about an operation that killed 83 people — as a problem.
Anthropic responded carefully: “We cannot comment on whether Claude, or any other AI model, was used for any specific operation, classified or otherwise.” But the damage to the relationship was done. The Pentagon began viewing Anthropic’s safety commitments not as responsible engineering but as potential insubordination.
Why This Matters for Ordinary People
You might think a contract dispute between an AI company and the Pentagon has nothing to do with your daily life. You’d be wrong. Here’s what’s actually at stake:
The Surveillance Infrastructure Is Already Built
US government agencies already purchase vast quantities of personal data from commercial data brokers — your location data, browsing history, purchase records, social media activity, and more. This data, collected from smartphone apps and advertising networks, provides surveillance capabilities that would require a warrant if obtained directly. By purchasing it on the open market, agencies argue they don’t need one.
Now imagine feeding all of that data into a system like Claude. An AI that can process billions of data points, identify patterns across millions of people, cross-reference communications with locations with financial transactions with social connections — all in real time. That’s not science fiction. That’s what removing the guardrails enables.
As Amodei himself wrote in his January 2026 essay “The Adolescence of Technology”: “A powerful AI looking across billions of conversations from millions of people could gauge public sentiment, detect pockets of disloyalty forming, and stamp them out before they grow.” He argued that large-scale AI-facilitated surveillance should be considered a crime against humanity.
Now the Pentagon is demanding that the man who wrote those words hand over his technology without restrictions.
If One Company Falls, They All Will
Anthropic isn’t the only AI company working with the military. Google, OpenAI, and Elon Musk’s xAI all operate on the Pentagon’s unclassified GenAI.mil platform. But Anthropic is the only one that has maintained explicit restrictions on autonomous weapons and domestic surveillance.
If the Pentagon successfully forces Anthropic to drop its guardrails — whether through contract pressure, supply chain designation, or the Defense Production Act — the message to every other AI company is unmistakable: resistance is not an option. No company will invest in safety measures that the government can simply order them to remove.
The Defense Production Act Precedent
If the government invokes the Defense Production Act against an AI company, it establishes that any technology company can be conscripted to build tools of surveillance and warfare, regardless of the company’s ethics, its users’ expectations, or its terms of service.
Your email provider, your cloud storage, your messaging app, your password manager — every technology company operates at the intersection of government power and user trust. The DPA precedent would tell all of them: when the government calls, you answer. Your users’ privacy is secondary to national security requirements that you don’t get to question.
AI Without Guardrails Is Surveillance Without Limits
The specific guardrails Anthropic is defending — no autonomous weapons, no mass surveillance — aren’t arbitrary. They represent the minimum viable ethical constraints on an extraordinarily powerful technology.
Remove the autonomous weapons restriction and you get AI systems making lethal decisions without human judgment, human compassion, or human accountability. Autonomous drone swarms that select their own targets. Intelligence systems that flag individuals for elimination based on pattern matching, not evidence.
Remove the mass surveillance restriction and you get an AI that can monitor every digital communication in the country simultaneously, flag dissent before it organizes, track every person’s movements and associations, and build predictive models of who might become a “problem” in the future. Not targeted investigation of suspected criminals — blanket surveillance of an entire population.
The Bigger Picture: Who Controls AI?
This confrontation is ultimately about a question that will define the next decade: who gets to decide how AI is used?
The Pentagon’s position is clear: the government decides, and companies comply. Hegseth’s demand that AI operate “without ideological constraints” reframes safety engineering as political bias — a rhetorical move that makes any ethical limitation on AI sound like partisan obstruction.
Anthropic’s position is that some uses of AI are too dangerous regardless of who’s asking, and that the company building the technology has not just a right but a responsibility to prevent catastrophic misuse.
For ordinary people, the stakes couldn’t be higher. If the government can force AI companies to remove safety guardrails, then the only protection between you and AI-powered mass surveillance is the government’s promise not to abuse the tools it demanded the right to use without restriction.
History suggests that promise is worth very little.
What You Can Do
The scale of this confrontation can feel paralyzing. A defense secretary issuing ultimatums to AI companies seems far removed from your daily life. But the surveillance infrastructure that this dispute is really about — the commercial data broker pipeline, the warrantless data purchases, the AI-powered analysis tools — that infrastructure runs on your data.
Every piece of data that exists about you is a data point that can be fed into these systems. The most effective defense isn’t political — it’s technical. Reduce the data that exists about you, and you reduce your exposure to whatever surveillance apparatus emerges from this fight.
-
Purge your data from brokers. Data brokers are the hidden pipeline that feeds your personal information to government agencies, advertisers, and anyone willing to pay. You can opt out — and it doesn’t have to cost a fortune. Our free Data Purge tool walks you through removing your information from major data brokers, step by step. The expensive “privacy services” charge hundreds of dollars for what you can do yourself.
-
Encrypt everything. Use Signal for messaging, Proton Mail or Tuta for email, and a privacy-respecting cloud service for file storage. Data that’s end-to-end encrypted can’t be read even if it’s intercepted or subpoenaed.
-
Minimize your digital footprint. Every account you create, every app you install, every service you sign up for is another data point. Delete accounts you don’t use. Remove apps that demand unnecessary permissions. Use cash when possible.
-
Choose your jurisdiction. Data stored in Switzerland has stronger legal protections than data stored in the United States. Jurisdiction isn’t just a corporate concern — it’s a personal privacy strategy.
-
Support companies that resist. Whether or not you agree with every decision Anthropic makes, the principle they’re defending — that technology companies can say no to government demands for unrestricted surveillance capabilities — is one that benefits everyone. Companies that maintain ethical red lines deserve your business over those that quietly comply with every request.
The Friday Deadline
As of this writing, Anthropic has shown no indication that it will capitulate. Amodei has reiterated the company’s red lines on autonomous weapons and mass surveillance, and sources familiar with the situation say the company has no plans to budge.
What happens Friday will set precedent for years. If Anthropic holds firm and survives the consequences, it proves that technology companies can resist government pressure to weaponize their products against civilians. If it folds — or is crushed — it proves they can’t.
Either way, your personal privacy strategy shouldn’t depend on the outcome of a corporate standoff. The tools to protect yourself exist today. Use them.
I Am NOT The Product was built for a world where you can’t trust governments or corporations to protect your privacy — because that world is the one we live in. Swiss-hosted, zero-knowledge encrypted, and designed so that surveillance is architecturally impossible. Start protecting your data today.