OpenAI and Google Employees Back Anthropic’s Pentagon Lawsuit

The Pentagon labeled Anthropic a supply chain risk, a designation typically reserved for foreign adversaries like Huawei. Within hours, Google’s chief scientist and dozens of OpenAI and Google employees filed a federal court brief to fight it.

On Monday, Anthropic sued the US Department of War over the blacklisting, and nearly 40 employees from rival companies filed an amicus brief, a court filing by non-parties with relevant expertise, the same day, crossing competitive lines rarely seen in the AI industry.

Rivals Unite in Federal Court

Among those who signed is Jeff Dean, Google’s chief scientist and head of Gemini. The signatories described themselves as engineers, researchers, scientists, and other professionals at leading U.S. AI companies, acting not on behalf of their employers but as individuals with direct technical knowledge of the systems at issue.

The brief drew a clear line between institutional and personal views. The signatories wrote:

“We submit this brief not as spokespeople for any single company, but in our individual capacities as professionals with direct knowledge of what these systems can and cannot do, and what is at stake when their deployment outpaces the legal and ethical frameworks designed to govern them.”

Amicus Brief Signatories, AI professionals at U.S. frontier AI laboratories

The brief frames the dispute not as one company’s contract fight but as a test of whether AI safety commitments can survive a procurement system that treats ethical constraints as commercial liabilities. The supply chain risk designation “is improper retaliation that harms the public interest,” the signatories wrote.

Despite coming from employees at Anthropic’s direct competitors, the brief asserts a shared professional consensus. According to the filing, the signatories acknowledged their diverse politics and philosophies yet agreed on a core stance: frontier AI systems are dangerous when deployed for domestic mass surveillance or autonomous lethal weapons without human oversight. They called for guardrails, whether technical or policy-based, to govern those uses.

That cross-partisan unity from two of Anthropic’s fiercest rivals gives the brief unusual weight as a signal of where the AI technical community stands on military AI ethics.

The Designation That Punishes Clients

The cross-industry solidarity centers on a federal procurement tool with no domestic precedent. The supply chain risk designation is typically reserved for foreign companies deemed a national security threat, such as Huawei. No domestic AI company has previously received this designation.

Moreover, the practical effect extends beyond Anthropic itself. The designation also blacklists other companies that use Anthropic products in Pentagon-related work, forcing any contractor using Claude to remove it in order to maintain military contracts. That downstream impact is why the signatories call it improper retaliation extending the harm beyond Anthropic to its enterprise clients and public-sector customers.

The designation followed a breakdown in talks. Negotiations broke down between Anthropic and the Trump administration after public disputes, while other AI companies signed contracts permitting unrestricted military use of their technology. Anthropic refused those terms, holding firm on its red lines against autonomous weapons and surveillance, specifically domestic mass surveillance and fully autonomous weapons capable of killing without human authorization. Those refusals, the designation implies, disqualify Anthropic as a trusted vendor.

Two Red Lines the Brief Defends

Beyond the procedural argument, the brief devotes substantial attention to the technical case for Anthropic’s two red lines, not as abstract ethical principles but as specific capabilities the signatories argue AI systems either cannot safely provide or should not provide.

Regarding surveillance, the brief argued that AI-enabled mass surveillance threatens democratic governance even when wielded by responsible actors. Surveillance data already exists across many government databases. What does not yet exist is the AI integration layer that could unify fragmented records into a unified, real-time surveillance apparatus combining face recognition, location history, transaction records, and social graphs.

Building on that, AI is the missing component that would make that infrastructure operational at scale, transforming fragmented data into a comprehensive monitoring system that no human team could build or manage manually. That distinction is why Anthropic’s refusal on surveillance is technically grounded rather than arbitrary.

Similarly, on lethal autonomous weapons, the brief contended that AI systems lack the precision and contextual judgment required for lethal targeting decisions. Current systems cannot reliably distinguish targets or weigh collateral costs the way a human operator can. The consequences of error in that context are irreversible. The signatories argued that removing human authorization from lethal decisions eliminates the one accountability mechanism that distinguishes lawful use of force from automated killing.

Taken together, both sections build toward the same conclusion: Anthropic’s refusals were not arbitrary corporate stances but reflect genuine capability limits and governance risks that the broader AI research community recognizes.

From Pentagon’s Chosen Partner to Blacklisted Vendor

However, the Pentagon’s reversal on Anthropic is sharp. Claude was the first AI model cleared for classified intelligence work, a milestone that made Anthropic one of the few AI companies trusted enough for sensitive government work. That clearance came after extensive vetting of Anthropic’s safety practices. Yet that partnership collapsed after Anthropic refused to extend it to surveillance and autonomous weapons applications.

The timing further underscores the contradiction at the center of the dispute. Hours after Defense Secretary Pete Hegseth announced the supply chain risk designation, the U.S. military reportedly used Claude in the very campaign that killed Iranian Supreme Leader Ayatollah Ali Khamenei, according to reporting cited in the amicus brief.

In effect, a model the Pentagon had just blacklisted was, within the same news cycle, an active tool in a consequential military operation – a striking illustration of the designation’s internal contradictions.

Prior Coverage and Context

Indeed, the Pentagon-Anthropic relationship had previously been one of the more notable defense AI partnerships in the industry. As WinBuzzer previously reported, the Pentagon AI contracts awarded to Google and Anthropic were part of a broader military AI push, cementing Anthropic’s standing as a preferred defense partner.

Furthermore, as we noted earlier, Anthropic had secured a major federal AI contract comparable in scale to OpenAI’s deal, positioning itself as a serious player in federal AI adoption alongside its larger rivals. That blacklisting therefore represents a sharp reversal of trajectory, from strategic partner to excluded vendor in a matter of months.

Implications Beyond One Contract

Thereversal from preferred partner to blacklisted vendor sets the stakes for what the amicus brief ultimately argues. The signatories concluded that Anthropic’s two red lines reflect concerns that are both technically grounded and broadly held across the AI research community, making the Pentagon’s retaliation a governance issue with implications extending well beyond one company’s contract.

The market signal, if the designation stands, is stark: AI companies that refuse to remove ethical constraints for military use will face procurement exclusion, while those willing to offer unrestricted terms gain competitive advantage in the defense market.

In contrast, the brief – signed by competitors who have no financial stake in Anthropic’s outcome – is a direct argument against that dynamic. The broader AI research community’s view, the brief argues, is that the technical concerns behind Anthropic’s red lines are legitimate and widely shared across the field, not the idiosyncratic limits of one cautious company.

The lawsuit is now before a federal court, with no ruling date set. The case has drawn attention beyond the immediate parties: experts at Oxford and defense policy analysts have warned that the designation sets a precedent with far-reaching consequences for how the U.S. government regulates AI procurement.

For the engineers, contractors, and enterprise customers caught between Pentagon contracts and safety commitments, the outcome will determine whether maintaining ethical limits on AI technology is a legally protected choice, or a commercially punished one.