Anthropic Said No to Mass Surveillance. The Industry Should Take Notes.

In late February 2026, something genuinely unprecedented happened in the AI industry. Anthropic, the company behind Claude, refused a direct ultimatum from the Pentagon to remove safety restrictions on its AI models. Defence Secretary Pete Hegseth gave CEO Dario Amodei a 72-hour deadline: allow unrestricted government use of Claude for mass surveillance and autonomous weapons, or face a federal blacklisting. Amodei said no.

AI ethics in practice means an AI company refusing to allow its technology to be used for mass domestic surveillance or fully autonomous weapons, even when threatened with losing a $200 million government contract and being designated a national security risk. This is not a philosophical debate any more. It happened, it cost Anthropic real money, and it has reshaped how the entire AI industry thinks about government partnerships. Here is exactly what happened, why it matters, and what came next.

What Did the Pentagon Want from Anthropic?

Anthropic had been providing Claude to US defence and intelligence agencies since late 2024 through a partnership with Palantir Technologies. Claude was the first frontier AI model deployed on classified Pentagon networks, available through Amazon's Top Secret Cloud. The contract was valued at up to $200 million and was announced in July 2025.

In early January 2026, Defence Secretary Pete Hegseth released an AI strategy memorandum directing that all Department of Defence AI contracts adopt standard "any lawful use" language, effectively requiring the removal of company-specific guardrails. The Pentagon demanded Anthropic drop two specific restrictions from its contract:

1. Mass domestic surveillance of American citizens -- Anthropic's contract explicitly prohibited using Claude for widespread monitoring of Americans.

2. Fully autonomous weapons -- Anthropic required human oversight in all lethal decision-making. No AI pulling the trigger without a human in the loop.

Pentagon spokesman Sean Parnell stated: "Our nation requires that our partners be willing to help our warfighters win in any fight."

How Did Anthropic Respond to the Ultimatum?

On 24 February 2026, Hegseth delivered the ultimatum face-to-face: comply by 5:01 PM on Friday 27 February, or face three consequences. First, cancellation of the $200 million contract. Second, designation as a "Supply-Chain Risk to National Security" -- a label historically reserved for US adversaries like Huawei, never before applied to an American company. Third, invocation of the Defence Production Act to force the removal of safeguards.

Amodei pointed out the contradiction: "Those latter two threats are inherently contradictory: one labels us a security risk; the other labels Claude as essential to national security."

On 26 February, Anthropic released its public statement. The Pentagon had presented new contract language framed as a compromise, but Anthropic said it "was paired with legalese that would allow those safeguards to be disregarded at will" and "made virtually no progress on preventing Claude's use for mass surveillance of Americans or in fully autonomous weapons."

Amodei's response was unequivocal: "We cannot in good conscience accede to their request."

Why Did Dario Amodei Refuse?

In an exclusive CBS News interview on 28 February, Amodei explained the reasoning behind what was, by any measure, an extraordinarily costly decision.

On mass surveillance, he made a point that deserves serious attention: "Mass domestic surveillance does not help the US catch up with its adversaries. It is an abuse of the government's authority, even where it is technically legal." He went further: "That actually is not illegal. It was just never useful before the era of AI. So there is this way in which domestic mass surveillance is getting ahead of the law."

This is the crucial insight. Under current US law, the government can purchase detailed records of Americans' movements, web browsing, financial transactions, and associations from commercial data brokers without obtaining a warrant. Before AI, analysing that data at scale was impractical. Now it is trivial. The law has not caught up with the technology.

On autonomous weapons, Amodei was equally direct: "The reliability is not there yet. We do not want to sell something that we do not think is reliable, and we do not want to sell something that could get our own people killed or that could get innocent people killed."

When asked whether he was being unpatriotic, his answer was memorable: "We are patriotic Americans. Everything we have done has been for the sake of this country. The red lines we have drawn we drew because we believe that crossing those red lines is contrary to American values." And then, perhaps the most important line: "Disagreeing with the government is the most American thing in the world."

What Happened After Anthropic Said No?

The retaliation was swift and severe. On 27 February, after the deadline passed, President Trump posted on Truth Social ordering every federal agency to immediately cease all use of Anthropic's technology, with a six-month phaseout. Hegseth designated Anthropic a Supply-Chain Risk to National Security. The State Department began replacing Claude with OpenAI's GPT-4.1 for its internal chatbot. Other agencies followed.

But here is where it gets interesting. The public response was the opposite of what the Pentagon expected.

Claude surged to number one on the US Apple App Store, up from 131st place. Anthropic recorded 503,424 downloads on 28 February alone, its largest single-day total ever. Free user signups rose 60%. The app crashed for four hours on 3 March from overwhelming demand. Bloomberg reported Anthropic neared a $20 billion revenue run rate, up from $9 billion at the end of 2025.

Meanwhile, #CancelChatGPT and #QuitGPT campaigns spread across social media, claiming over 1.5 million participants. ChatGPT uninstalls spiked 295% in a single day. The direct financial impact of the Pentagon ban was limited: Bloomberg Government records showed the Pentagon had only paid $2 million to Anthropic in the prior year, against a company generating billions in commercial revenue.

How Did OpenAI Respond to This Situation?

Hours after Trump's ban on Anthropic, OpenAI announced it had struck a deal with the Pentagon to fill the gap. The timing was widely criticised. Sam Altman initially told staff in an internal memo that he wanted to "try to help de-escalate things" and claimed OpenAI shared Anthropic's red lines on mass surveillance and autonomous weapons.

But the backlash was immediate. 573 Google employees and 93 OpenAI employees signed an open letter titled "We Will Not Be Divided", which stated: "The Pentagon is negotiating with Google and OpenAI to try to get them to agree to what Anthropic has refused. They are trying to divide each company with fear that the other will give in. That strategy only works if none of us know where the others stand." By 3 March, it had nearly 900 signatures.

Noam Brown, a prominent OpenAI researcher, publicly objected to the original contract language, saying it left "legitimate questions unanswered, particularly around new ways AI could potentially enable lawful surveillance."

By 3 March, Altman admitted the deal looked "opportunistic and sloppy" and announced contract amendments. The revised contract added explicit bans on mass domestic surveillance, excluded intelligence agencies (NSA, NGA, DIA), added protections for commercially acquired data, and prohibited autonomous weapons.

Why Are Legal Experts Still Concerned About OpenAI's Deal?

Despite the amendments, legal experts flagged significant weaknesses. Jessica Tillipman, Associate Dean for Government Procurement Law at George Washington University, wrote that OpenAI's contract "does not give OpenAI an Anthropic-style, free-standing right to prohibit otherwise-lawful government use." It simply states the Pentagon cannot use OpenAI's technology to break laws and policies as they are stated today. But those laws and policies can change.

Samir Jain from the Center for Democracy and Technology warned: "Right now, under US law, it is lawful for government authorities to buy commercially available information and use AI to analyse it at scale, in effect engaging in mass surveillance of Americans. Nothing in what they have released would prevent those policies from being changed."

This is the fundamental difference. Anthropic demanded a contractual right to prohibit these uses regardless of legality. OpenAI's contract merely references existing legal frameworks, which the government can change unilaterally. One is a genuine red line. The other is a guardrail that can be moved.

What Does This Mean for the AI Industry?

This episode has set a precedent that will shape the AI industry for years. Several things are now clear.

AI companies can say no and survive. Anthropic lost its government contract but gained enormous public goodwill, record downloads, and commercial momentum. The financial impact was minimal compared to its commercial revenue. This matters because it removes the excuse that companies "have to" comply with government demands to survive.

The law has not caught up with AI capabilities. The fact that mass surveillance of Americans through commercially acquired data is technically legal but was never practically possible before AI is a regulatory gap that urgently needs closing. Until it is, the only protection citizens have is the ethical stance of the companies building these tools.

Employee pressure works. The open letter from Google and OpenAI employees, combined with public backlash, forced OpenAI to amend its contract within days. The "We Will Not Be Divided" letter demonstrated that workers across competing companies can act collectively on ethical issues.

Consumers vote with their wallets. The #QuitGPT movement and the surge in Claude downloads show that ethics is a competitive advantage, not a liability. In a market where AI products are increasingly similar in capability, trust becomes the differentiator.

For businesses choosing AI partners, this raises important questions. Who built the tools you rely on? What are their values? What would they do if a government asked them to surveil your customers? These are no longer hypothetical questions. They are business decisions with real consequences. If you are evaluating AI tools for your business, our AI automation services can help you choose tools that align with your values and your customers' privacy expectations.

FAQ: Anthropic, the Pentagon, and AI Ethics

Q: Did Anthropic refuse to work with the military entirely?

No. Anthropic made clear it supports all lawful uses of AI for national security. Its two red lines were specifically mass domestic surveillance of American citizens and fully autonomous weapons without human oversight. It was willing to continue all other defence work.

Q: Is mass surveillance of Americans actually legal?

Currently, yes, in many practical respects. The US government can legally purchase commercially available data, including location tracking, web browsing history, and financial records, from data brokers without a warrant. AI makes it possible to analyse this data at population scale for the first time, creating a surveillance capability that existing privacy laws were not designed to address.

Q: What happened to Claude's availability after the ban?

Claude remains fully available to consumers and businesses worldwide. The ban only affects US federal government agencies. Commercial availability was unaffected, and demand actually surged, with Claude reaching number one on the App Store and Anthropic approaching a $20 billion revenue run rate.

Q: Is OpenAI now providing AI for mass surveillance?

OpenAI's amended contract includes explicit bans on mass domestic surveillance and excludes intelligence agencies from access. However, legal experts note these protections reference existing law, which the government can change, rather than providing the kind of absolute contractual prohibition Anthropic demanded.

Q: How does this affect businesses choosing AI tools?

It highlights the importance of understanding your AI provider's values and data policies. For businesses handling sensitive customer data, the ethical stance of your AI partner matters. At Northern Codes, we help businesses choose and implement AI tools with privacy and ethics built in. Get in touch for a free consultation.

The Bottom Line

Dario Amodei and Anthropic did something that very few companies in history have done: they told the most powerful government on Earth "no" and accepted the consequences. They lost a $200 million contract, got blacklisted from federal work, and were publicly attacked by the President. And they came out stronger for it, with record downloads, surging revenue, and a level of public trust that money cannot buy.

The AI industry is at a crossroads. The technology to surveil entire populations now exists. The legal frameworks to prevent it do not. Until the law catches up, the only thing standing between mass surveillance and the public is the integrity of the people building these systems. Anthropic drew the line. The rest of the industry should take notes.

If you are thinking about how AI can help your business while respecting your customers' privacy, explore our services or request a free quote. We build AI solutions that put your values first.

Sources:

Anthropic - Statement from Dario Amodei on Department of War discussions | CBS News - Anthropic CEO sticking to AI red lines | NPR - OpenAI announces Pentagon deal after Trump bans Anthropic | CNBC - Altman admits deal looked opportunistic and sloppy | TechCrunch - Employees at Google and OpenAI support Anthropic in open letter | OpenAI - Our agreement with the Department of War | Bloomberg - Anthropic nears $20 billion revenue run rate | MIT Technology Review - OpenAI's compromise is what Anthropic feared