Back to Resources
AI Governance12 min read
The Anthropic Defence Contract Controversy: What It Means for AI Governance
How Anthropic's refusal of military surveillance contracts, the supply chain risk designation, and the user backlash reshaped the AI industry.
The Original Deal: Anthropic and the Pentagon
In July 2025, the Pentagon's Chief Digital and Artificial Intelligence Office (CDAO) awarded Anthropic a two-year, $200 million prototype contract to deploy its Claude AI model across U.S. defence workflows. Anthropic had already been working with the government since June 2024, becoming the first frontier AI company to deploy models on classified networks. Claude was integrated into mission workflows with partners like Palantir, used for intelligence analysis, operational planning, cyber operations, and modelling and simulation.
The contract came with two specific conditions that the Pentagon initially accepted. Anthropic's acceptable use policy prohibited the use of Claude for mass domestic surveillance of Americans and for fully autonomous weapons systems — those capable of selecting and engaging targets without human intervention. Similar contracts worth up to $200 million each were simultaneously awarded to OpenAI, Google DeepMind, and xAI, bringing the total potential value to $800 million.
For businesses evaluating AI providers, this context matters. Anthropic was not an outsider making ideological demands — it was the Pentagon's preferred AI vendor, widely regarded as producing superior results for government use cases. Federal employees found Claude's step-by-step reasoning particularly valuable for agencies requiring strong auditing and verification.
The 'Any Lawful Use' Demand and Anthropic's Refusal
The situation changed in January 2026 when the Pentagon released a new AI Strategy memorandum directing all DoD AI contracts to incorporate a standard "any lawful use" clause within 180 days. This represented a broader push to accelerate America's military AI dominance to outpace China — even if safety guardrails were not fully established.
Defence Secretary Pete Hegseth told Anthropic CEO Dario Amodei that if the company did not allow Claude to be used "for all lawful purposes," the Pentagon would cancel its contract. The Pentagon's proposed compromise language, Anthropic said, "was paired with legalese that would allow those safeguards to be disregarded at will."
On February 26, Amodei publicly rejected the Pentagon's final offer. His statement was unequivocal: "Their threats do not change our position: we cannot in good conscience accede to their request." He argued that frontier AI systems are "simply not reliable enough to power fully autonomous weapons" and that mass surveillance capabilities are "getting ahead of the law." The Pentagon CTO responded that it was "not democratic" for Anthropic to limit military use of its technology.
White House AI czar David Sacks had already been publicly attacking Anthropic, calling it "woke AI" and accusing the company of pursuing "a sophisticated regulatory capture strategy based on fearmongering." The ideological framing was deliberate: Sacks accused Anthropic of trying to "backdoor Woke AI and other AI regulations through Blue states like California."
The Pentagon set a deadline of 5:01 PM on Friday, February 27 for Anthropic to accept the government's terms. Anthropic did not agree.
The Government Strikes Back: Ban and Supply Chain Risk
On February 27, 2026, President Trump ordered all federal agencies to cease using Anthropic products, with a six-month phase-out period. Trump publicly labelled Anthropic a "radical left, woke" AI company. Hegseth declared: "America's warfighters will never be held hostage by the ideological whims of Big Tech."
The retribution escalated further. On March 3-5, the DoD formally designated Anthropic a "supply chain risk" — a label historically reserved exclusively for foreign adversaries like Huawei and Kaspersky. Anthropic became the only American company ever to receive this designation publicly.
The practical consequences are severe:
• Defence contractors must now certify they do not use Anthropic's models in any Pentagon-related work
• Enterprise customers with Pentagon contracts face pressure to cut ties with Anthropic across their entire operations
• Federal agencies including HHS, Treasury, and State have confirmed they are transitioning off Claude
• Revenue impact could reach multiple billions of dollars in 2026, according to industry analysts
As Mayer Brown's legal analysis noted, the designation's scope extends far beyond Anthropic's direct government contracts — it threatens the company's entire enterprise business by forcing any government-adjacent customer to avoid Anthropic products entirely.
OpenAI Steps In — Then Stumbles
Hours after Trump's ban, OpenAI CEO Sam Altman announced his company had secured the Pentagon contract to replace Anthropic on classified networks. OpenAI accepted the "any lawful use" clause while negotiating stated safeguards: no mass domestic surveillance, no directing autonomous weapons, cloud-only deployments, and security-cleared OpenAI engineers embedded within the Pentagon.
The timing drew immediate criticism. Altman admitted on March 3 that the deal "looked opportunistic and sloppy" and announced contract amendments. The Electronic Frontier Foundation was considerably harsher, calling OpenAI's safeguard language "weasel words" that would not prevent AI-powered surveillance. The full contract text has never been released publicly.
The fallout hit OpenAI internally. CNN reported that employees were "fuming" about leadership's handling of the negotiations, with many saying they "really respect" Anthropic's stance. Caitlin Kalinowski, OpenAI's hardware leader and former Meta AR executive, resigned citing ethical concerns. Activists chalked messages on sidewalks outside OpenAI's San Francisco offices: "Where are your redlines?" and "You must speak up."
The Intercept summarised OpenAI's position on surveillance and autonomous weapons bluntly: "You're going to have to trust us." For enterprise buyers weighing AI providers, the contrast in corporate governance could not have been starker.
The Public Responds: Claude Surges to Number One
Consumer reaction was swift and decisive. On approximately February 28, one day after the Pentagon terminated its relationship with Anthropic, Claude surged to the No. 1 spot on the Apple App Store, overtaking ChatGPT for the first time.
The numbers tell a striking story:
• Claude jumped from No. 131 on January 30 to No. 1 by late February
• Daily downloads hit 149,000 versus ChatGPT's 124,000 by March 2
• Daily active users reached 11.3 million, up 183% from roughly 4 million at the start of 2026
• Paid subscribers doubled
• Web traffic rose 43% month-over-month in February and 297.7% year-over-year
• Claude became the top AI app in more than 20 countries
Social media flooded with screenshots of users cancelling ChatGPT subscriptions and signing up for Claude. Pop star Katy Perry posted a screenshot of Claude's Pro page with a red heart. Chalk messages saying "Thank you" appeared on sidewalks outside Anthropic's San Francisco offices.
TechCrunch reported that the consumer growth surge continued well into March. While ChatGPT still holds a commanding lead in total users (250.5 million daily active users versus Claude's 11.3 million), the trajectory shift is unmistakable — and it happened because a company refused to compromise on stated principles.
For business leaders evaluating AI vendors, this episode demonstrated something important: principled governance is not just an ethical nicety — it can be a powerful market differentiator.
Anthropic's Legal Challenge and the Case for Injunctive Relief
On March 9, 2026, Anthropic filed two federal lawsuits against the Trump administration — one in the U.S. District Court for the Northern District of California and another in the federal appeals court in Washington, D.C.
Anthropic's legal arguments centre on three claims:
• First Amendment retaliation — the supply chain risk designation punishes Anthropic for its public advocacy on AI safety, not for any genuine national security concern
• Statutory overreach — the supply chain risk label was designed by Congress for foreign adversary threats, not domestic contract disputes; the DoD failed to follow legally required procedures including risk assessment, notification, and Congressional notice
• Irreparable harm — Anthropic's CFO stated in a filing that the company risks losing "hundreds of millions" in 2026 revenue
The DOJ filed its opposition on March 17-18, arguing the designation is "lawful and reasonable" and that Anthropic's refusal to accept contract terms constitutes "conduct, not speech." The DOD also raised an unusual concern: that Anthropic might "attempt to disable its technology or preemptively alter the behavior of its model" during critical military engagements.
The support for Anthropic has been broad and bipartisan. Nearly 150 retired federal and state judges — appointed by both Republican and Democratic presidents — filed an amicus brief. They were joined by Microsoft, dozens of OpenAI and Google DeepMind researchers (in personal capacities), major tech industry groups, and former senior national security officials. Even Sam Altman publicly stated he does not believe Anthropic should be labelled a supply chain risk. Lawfare's analysis argued the designation "won't survive first contact with the legal system."
A preliminary injunction hearing is scheduled for March 24, 2026 before Judge Rita F. Lin in San Francisco.
What This Means for AI Governance
The Anthropic-Pentagon dispute is more than a contract disagreement. According to Oxford University's analysis, it "reflects governance failures — with consequences that extend well beyond Washington." Chatham House concluded that the feud "reveals the limits of AI governance" — contractual mechanisms between individual companies and government agencies are no substitute for proper institutional frameworks.
Several structural observations emerge:
• Private companies are setting guardrails that governments haven't. The EFF pointed out that privacy protections "shouldn't depend on the decisions of a few powerful people" — whether those people are tech CEOs or defence secretaries. Congress, not private companies, should be defining the rules.
• The power dynamic has shifted. AI capability is now concentrated in commercial firms rather than government laboratories, a departure from decades of defence innovation. This gives private companies leverage they have never previously held in the military procurement process.
• The precedent extends globally. NYU Stern's Center for Business & Human Rights analysed "The Cost of Conscience" — how this dispute sets expectations for every AI company worldwide that engages with government clients.
• Internal dissent is a leading indicator. More than 100 Google employees sent an internal letter asking for similar limits on how Gemini models are used by the military. The tech industry broadly rallied behind Anthropic's position, suggesting that the "any lawful use" approach may struggle to attract top AI talent.
The Council on Foreign Relations characterised the situation as a dispute "about politics and personalities" that is "masquerading as a policy dispute." Foreign Policy called it "a bad sign" for the broader defence-tech ecosystem.
Implications for Businesses Choosing an AI Partner
For organisations evaluating enterprise AI providers, the Anthropic-Pentagon saga surfaces questions that go well beyond model benchmarks:
Governance matters commercially. Anthropic's consumer surge proved that principled governance can drive growth. Businesses increasingly prefer vendors whose values they can articulate to their own customers and stakeholders. The companies that stood firm on stated principles attracted users; the one that rushed to fill the gap had to publicly apologise within a week.
Supply chain risk creates real exposure. If your organisation has any relationship with U.S. defence contracting — even indirectly through vendors or partners — the supply chain risk designation is a live compliance issue. Mayer Brown's guidance for government contractors details the certification requirements now in effect.
The legal landscape is unsettled. The March 24 hearing will provide the first judicial signal. A ruling for Anthropic would suggest the government overreached; a ruling for the DoD would empower its hardline stance. Either way, this case could set landmark precedent for how AI companies engage with government clients — and what rights they retain when they disagree.
Corporate speech protections are being tested. Anthropic's First Amendment argument — that the government retaliated against it for publicly advocating AI safety positions — has implications for any technology company that takes public stances on responsible AI. The outcome will signal whether companies can maintain ethical red lines without risking government retribution.
For businesses that value transparent, accountable AI governance in their technology partners, this controversy has made the differences between providers impossible to ignore. The question is no longer just "which model performs best" — it is "which company will stand behind its principles when the pressure is real?"
This situation continues to develop. A hearing on Anthropic's preliminary injunction is set for March 24, 2026, and the outcome will shape the AI governance landscape for years to come.
PreviousClaude vs ChatGPT for Business (2026): A Data-Driven ComparisonNext How to Migrate from ChatGPT to Claude (Complete 2026 Guide)
Want to Apply This to Your Business?
We're a Singapore AI development and automation agency. Let's discuss how we can help solve your specific challenges.