More than 30 engineers and researchers from OpenAI and Google DeepMind have signed a public letter in defense of Anthropic in its legal battle against the U.S. Department of Defense (DOD). This isn’t a coordinated corporate effort; it’s personal. And that certainly makes it unusual in Silicon Valley, where loyalty to one’s employer usually takes precedence over any industry principle.
The DOD's lawsuit against Anthropic, alleging a breach of contract, is generating an unexpected collateral effect. For the first time in years, employees across the three major AI labs are finding common ground beyond their commercial rivalries. The implicit message is clear: what happens to Anthropic affects us all. This means that not only are government contracts at stake, but also the operational boundaries of how AI companies can work with the government without compromising their core principles.
The Letter OpenAI and Google Didn’t Want (But Couldn’t Avoid)
The statement was published last Tuesday in an open letter coordinated by an informal group of AI safety researchers. It includes recognizable names: three team leads from OpenAI, five senior researchers from DeepMind, and about twenty other technical professionals with notable standing in their organizations. What’s interesting is that none of them spoke on behalf of their companies. They all signed with their names.
The content is straightforward: Anthropic has the right to set its own red lines regarding the military uses of its models. The government cannot force a private company to develop capabilities it deems ethically problematic. And, crucially, compelling this collaboration through legal action sets a dangerous precedent for the entire industry.
What’s intriguing isn’t so much the content, which is predictable in the realm of AI safety, but rather the timing and who has signed it. OpenAI has been negotiating its own contracts with the DOD for months. Google, on the other hand, has a complicated history with the Pentagon following the internal rebellion over Project Maven in 2018. The fact that employees from both companies are publicly defending a direct competitor speaks to a shared anxiety: if the government can force Anthropic, it can force us all.
The Context That Explains the Urgency
The DOD's lawsuit against Anthropic stems from a contract signed in 2024 to develop intelligence analysis capabilities using Claude. According to leaked court documents, Anthropic refused to expand the contract’s scope to include targeting and real-time tactical analysis applications, citing its Acceptable Use Policies. The DOD argues there has been a breach of contract. Anthropic counters that it never agreed to those terms and that the government is attempting to reinterpret the original agreement.
Legally, this is a complex case. But, honestly, it’s a technical bombshell: it forces the industry to define where "intelligence analysis" ends and "direct military application" begins. And truthfully, nobody has that line clear.
Why This Time Is Different (And Why It Matters)
The AI safety community has debated dual-use concerns for years in academic conferences and papers that few people read. However, a lawsuit from the U.S. government is a different story. This turns a philosophical debate into a legal issue with far more tangible contractual consequences.
The signatories of the letter argue three technical points that, in my view, go beyond mere rhetoric:
1. Contractual Precedent: If the DOD wins, any government contract with open clauses could be reinterpreted in the future to include military applications that weren't initially specified. This would mathematically complicate the relationship of any AI startup with federal contracts.
2. Talent Drain: Most top AI researchers have limitless job options. Many chose to work at Anthropic, OpenAI, or DeepMind precisely because of their public commitments to AI safety. Forcing these companies to compromise those principles through legal pressure not only affects the company but also has repercussions on the talent pool in the sector.
3. Global Fragmentation: If U.S. AI companies are compelled to cooperate with military applications against their better judgment, top researchers in Europe and Asia will have a clear incentive to work in labs outside U.S. jurisdiction. Both China and the EU are observing this case with great interest.
What They Don’t Say (But Everyone Understands)
None of the signatories explicitly mention the word "precedent," but the memory of Project Maven looms over every line. In 2018, over 4,000 Google employees signed an internal letter opposing the Pentagon’s image analysis contract. Google ultimately did not renew that contract and published its AI Principles. It was a victory for employees, but also a defining moment: it showed that in this industry, talent holds significant power.
The context this time is different. This isn’t an internal protest, but a public defense of a competitor. And that changes the dynamic: they are not fighting for a company’s values, but for the operational rules of the entire industry.
The Game Within the Game: What the Signatories Gain and Risk
Signing this letter comes with its costs. None of the 30-plus employees signed anonymously, meaning they are assuming real professional exposure. In organizations where government contracts represent significant revenue, especially for OpenAI after its agreements with various federal agencies, taking a public stance against the DOD is hardly a career booster.
However, the risk-benefit calculation seems clear to them: the individual risk is less than the collective risk of normalizing legal coercion to force military applications of AI models.
There's also a generational component. Most of the signatories are between 28 and 42 years old. They are millennials and Gen Z who grew up witnessing the consequences of wars in Iraq and Afghanistan, as well as Snowden's whistleblowing and the revelations of Cambridge Analytica. In my experience, for this generation, "working for the government" doesn’t carry the same automatic moral weight as it did for those who built companies like Google or Facebook in the 2000s.
Corporate Silence Is Also a Signal
Neither OpenAI, Google, nor Anthropic has issued official statements regarding the letter. Their silence is strategic: it allows employees to voice their opinions without compromising the companies' contractual negotiations. However, this silence is also revealing. It suggests that, internally, leadership does not fundamentally disagree with their employees. They are simply more cautious about how to express it publicly.
Sam Altman has not commented on the matter. Demis Hassabis hasn’t either. Dario Amodei, CEO of Anthropic, has only thanked the community for its support. Everything is measured. No one wants to give the DOD more ammunition for the legal case.
What This Reveals About the Current Moment in AI
This situation wouldn’t exist if we were in 2021. Back then, AI labs were competing for talent, funding, and resources. Now, in addition to that, they are also competing for moral legitimacy in a context where governments are starting to treat them as potential defense contractors.
The public debate around AI has evolved from "Can machines think?" to "Who controls the machines that think?" And that second question is key, fundamentally political. The DOD's lawsuit is merely the first visible case of a tension that has been developing for years: can private AI companies set their own ethical lines when developing general-purpose technology with obvious military applications?
The outcome of this legal case will define much more than a specific contract. It will determine whether companies like OpenAI, Anthropic, or Google can continue to operate with relative autonomy over their technical and ethical principles, or if they will eventually have to choose between complying with the government or losing access to the U.S. market.
And that choice, if it comes, will be far more complex than a simple debate over AI ethics. It will be about who holds the real power in the 21st-century tech ecosystem.
The question that no one dares to ask publicly, but everyone asks privately is: how far are governments willing to go to secure a strategic advantage in AI? And more importantly: how far are those who build those systems willing to resist? This letter from 30 employees is just the beginning of a much longer conversation.
Does your company have a clear red line on what to do (and what not to do) with AI? If not, it will need one soon.