Anthropic is getting backup from an unlikely source: employees at its fiercest competitors.
A lawsuit filed by the U.S. Department of Defense against Anthropic — challenging the AI company’s refusal to strip ethical guardrails from its Claude models deployed in military contexts — has triggered an unusual wave of cross-company solidarity. Engineers, researchers, and policy staffers at OpenAI and Google DeepMind have signed an open letter defending Anthropic’s right to maintain its responsible use policies, even when contracting with federal agencies. The letter, which circulated internally before going public, had gathered over 400 signatures from employees across the three companies as of March 9, 2026, according to TechCrunch.
The core dispute is straightforward. The DOD contracted Anthropic in late 2025 to provide Claude-based AI systems for logistics, intelligence analysis, and operational planning. But the contract included Anthropic’s standard Acceptable Use Policy, which restricts certain military applications — including autonomous targeting and lethal decision-making support. The Pentagon argues those restrictions violate the terms of the federal procurement agreement and effectively give a private company veto power over national defense capabilities.
Anthropic isn’t budging.
“We built these safety constraints into our systems because they reflect our core technical and ethical commitments,” Anthropic’s head of policy told reporters. “Removing them isn’t a toggle you flip. It’s a fundamental change to how the model operates.” The company has framed its position not just as a moral stance but as a technical one — arguing that stripping safety layers could introduce unpredictable behavior in high-stakes military environments.
And that’s where the competitor support gets interesting. The open letter from OpenAI and Google employees doesn’t just express philosophical agreement. It makes a pointed technical argument: that weakening safety fine-tuning in frontier models for any single customer creates systemic risk across the broader AI industry. “If the government can compel one company to remove alignment guardrails under contract law, every AI provider working with federal agencies faces the same precedent,” the letter reads.
Some context matters here. OpenAI has its own Pentagon contracts. So does Google, which weathered the infamous Project Maven controversy back in 2018 before eventually re-engaging with defense work through Google Cloud. Neither company has publicly endorsed the letter — their employees signed as individuals, not representatives. But the sheer number of signatories from rival firms signals something deeper than personal conviction. It suggests a shared anxiety about where government AI procurement is headed.
The DOD’s legal theory is aggressive. Federal prosecutors argue that once Anthropic accepted the contract and the associated funding — reportedly worth over $800 million — the company became bound to deliver the full capability of its models without self-imposed restrictions that weren’t negotiated as formal carve-outs. In other words: you took the money, now deliver. Legal experts are split on whether this holds up. Government contracting law does generally favor the procuring agency’s requirements, but there’s limited precedent for compelling a software vendor to modify the fundamental behavior of an AI system post-deployment.
“This case could define the boundaries between private AI governance and federal authority for a generation,” said Stanford HAI’s Marietje Schaake in comments reported by TechCrunch.
She’s probably right.
The timing compounds the tension. Congress is actively debating the AI Accountability Act of 2026, which would impose mandatory safety standards on frontier AI systems. If passed, it could actually support Anthropic’s position — codifying the kind of restrictions the company already applies voluntarily. But the bill faces fierce opposition from defense hawks who argue it would hamstring American military AI capabilities relative to China’s rapidly advancing programs.
Meanwhile, Anthropic’s competitors are walking a tightrope. OpenAI CEO Sam Altman has not commented publicly on the lawsuit, though sources familiar with the company’s thinking told TechCrunch that leadership is “watching closely.” Google’s position is similarly guarded. Both companies have their own acceptable use policies for military applications, and both would face identical legal exposure if the DOD prevails against Anthropic.
The employee letter makes this explicit. “Today it’s Anthropic. Tomorrow it’s any of us,” one passage reads. Not subtle.
There’s a business dimension too. Anthropic has positioned its safety-first approach as a competitive differentiator, particularly with enterprise and government clients in allied nations where AI ethics requirements are strict. A court ruling forcing the company to abandon those principles wouldn’t just affect one contract — it could undermine Anthropic’s brand proposition globally. European and Australian defense partners have specifically cited Anthropic’s safety commitments as a reason for preferring Claude over alternatives, according to procurement documents reviewed by TechCrunch.
So what happens next? The case is currently in the U.S. Court of Federal Claims. Anthropic has filed a motion to dismiss, arguing the DOD’s interpretation of the contract is overbroad and that compelling modifications to AI safety systems raises First Amendment concerns — a novel and untested legal theory. The DOD has opposed the motion. Oral arguments are expected in April.
No matter the outcome, this lawsuit has already changed the conversation. The fact that hundreds of employees at competing companies would publicly align themselves with Anthropic — against a government client that writes checks to all of them — tells you how seriously the technical AI community takes this fight. It’s not about one contract or one company. It’s about whether the companies building the most powerful AI systems get to set boundaries on how those systems are used, or whether that power belongs entirely to whoever’s paying.
That question doesn’t have an easy answer. But it now has a courtroom.
WebProNews is an iEntry Publication