30+ OpenAI and Google Employees Back Anthropic in Pentagon AI Ethics Lawsuit

On March 9th, more than 30 employees from OpenAI and Google DeepMind signed an amicus brief supporting Anthropic’s lawsuit against the Department of Defense. Among the signatories: Google DeepMind Chief Scientist Jeff Dean.

The backdrop: the Pentagon designated Anthropic a supply-chain risk after Anthropic refused to allow its models to be used for mass surveillance of Americans or autonomous weapons targeting without human oversight. The DOD then signed a competing contract with OpenAI.

Employees from the company that got the competing contract are now publicly defending the company that refused. That’s the story.

What Anthropic Refused to Do

Anthropic’s lawsuit details the specific use cases it declined to support:

• Mass surveillance of US persons using AI-generated behavioral profiles
• Autonomous weapons targeting systems with no human in the loop for lethal decisions
• Predictive policing systems that use AI outputs as the primary basis for detention or action

Anthropic’s acceptable use policy prohibits these applications. When the DOD required them as contract terms, Anthropic said no. The Pentagon classified this as a supply-chain risk — meaning Anthropic’s refusal to waive its safety policies became grounds for exclusion from federal contracts.

Why Competitors Are Defending Them

The amicus brief from OpenAI and Google DeepMind employees makes an argument that cuts across competitive lines: if the government can penalize AI companies for maintaining ethical use limits, the structural incentive shifts toward removing those limits to stay competitive in federal procurement.

That’s a race to the bottom that’s bad for the industry regardless of which company wins the contracts.

Jeff Dean’s involvement is notable because Google DeepMind has its own substantial government and defense contracts. Signing onto a brief defending a competitor’s right to refuse certain military use cases is a signal about where at least some of the industry’s technical leadership wants to draw lines — even when it costs business.

What Happened After Anthropic Was Blacklisted

The DOD moved quickly. Within weeks of designating Anthropic a supply-chain risk, it signed a new AI contract with OpenAI — the same company whose employees are now backing Anthropic in court.

That sequence matters for understanding the incentive structure. The message the DOD sent: decline our terms and we go to your competitor. The message the amicus brief sends back: we don’t think that’s an acceptable precedent.

What This Means for Builders Working With Government

If you’re building AI products for public-sector clients or considering federal contracts, this case is the most important thing happening in AI procurement right now.

The outcome will determine:

• Whether acceptable-use clauses in federal AI contracts can be enforced by contractors
• Whether safety policies that limit certain government use cases constitute grounds for exclusion
• What “supply-chain risk” can legally mean in the context of AI vendor selection

The practical advice until this resolves: if you’re in federal AI sales, have explicit conversations with legal about which use cases your product’s acceptable-use policy prohibits, and make sure your contract review process flags any DOD requirements that conflict with those limits before you’re in Anthropic’s position.

---

Source: TechCrunch — “OpenAI and Google employees rush to Anthropic’s defense in DOD lawsuit” | Fortune, March 10, 2026