The Pentagon chose capability and speed over constraint.
Pentagon partnered with SpaceX, OpenAI, Google, NVIDIA, Reflection, Microsoft, and AWS to accelerate AI-driven military capabilities and maintain U.S. technological superiority. Anthropic rejected the contract over refusal to accept surveillance and lethal autonomous weapons clauses, creating a rare ethical stand against Pentagon demands.
- Pentagon signed AI contracts with SpaceX, OpenAI, Google, NVIDIA, Reflection, Microsoft, and AWS
- Anthropic rejected Pentagon terms requiring removal of restrictions on mass surveillance and lethal autonomous weapons
- Over 600 Google employees formally objected to the company's military AI partnerships
- Anthropic's technology had been deployed in the Iran war and January 2026 Venezuela operation before the company severed ties
The U.S. Department of Defense signed agreements with seven major tech companies to integrate AI into military operations, notably excluding Anthropic due to ethical disagreements over surveillance and autonomous weapons.
The Pentagon has locked in artificial intelligence partnerships with seven major technology companies—SpaceX, OpenAI, Google, NVIDIA, Reflection, Microsoft, and Amazon Web Services—in a sweeping move to embed machine learning across military operations. The announcement, made by the Department of Defense under leadership that views AI dominance as essential to national security, marks a significant acceleration in the militarization of advanced technology. One company's absence from the list speaks as loudly as the seven names that made it: Anthropic, the AI developer that walked away from Pentagon contracts over fundamental disagreements about what its technology should be used for.
The core disagreement centers on two capabilities the Pentagon wanted and Anthropic refused to enable. The company rejected contract terms that would have allowed its AI systems to power mass surveillance operations and to support lethal autonomous weapons—machines that could select and fire on targets without human intervention. When Anthropic declined to remove those restrictions, the Pentagon moved forward without them. The rupture happened early in the year, and since then, other firms have stepped in to fill the space Anthropic vacated, operating within what the Pentagon describes as clearly regulated environments.
Why does the Pentagon believe this matters so urgently? The department frames AI as a force multiplier for decision-making under pressure. Military commanders face torrents of data from sensors, satellites, and field reports. AI systems can synthesize that information at machine speed and recommend actions in real time. In complex theaters—where multiple nations and armed groups operate simultaneously, where threats emerge faster than traditional analysis can process them—that speed advantage translates into strategic advantage. The Pentagon's official language emphasizes that these partnerships will "accelerate the transformation toward establishing United States Armed Forces with AI at the center of how they fight." The timing is not incidental. The agreements arrive against the backdrop of recent military interventions: a war with Iran that is currently in a ceasefire, and a military operation in Venezuela in January aimed at capturing President Nicolás Maduro. In both cases, according to experts, Anthropic's technology had been deployed before the company severed ties with the Pentagon.
The seven companies that signed on represent the architecture of American technological power. OpenAI brings large language models and reasoning systems. Google contributes its deep learning infrastructure and data processing capabilities. Microsoft and AWS offer cloud computing at scale. NVIDIA supplies the specialized chips that make AI training and inference possible. SpaceX and Reflection round out the roster with their own technical specialties. Together, they form a consortium designed to ensure that American military systems operate on American-built AI, reducing dependence on foreign technology and consolidating control over the most sensitive applications.
But the announcement has fractured Silicon Valley along lines that were already visible before the Pentagon made its move. At Google, more than 600 employees have formally objected to the company's military contracts and demanded that leadership halt new defense agreements. The internal resistance reflects a broader tension in the technology industry: some companies see government partnership as inevitable and profitable, while others—Anthropic most visibly—have chosen to build their business around ethical constraints, even when those constraints cost them revenue and influence. Anthropic's position is not that AI should never be used in defense; it is that certain applications cross a line that the company will not cross, and that the Pentagon's willingness to move forward without them suggests the Pentagon does not share that boundary.
What happens next will likely depend on how effectively these seven partnerships deliver on their promises. If AI-driven military systems prove transformative—if they genuinely accelerate decision-making and reduce casualties among American forces—the Pentagon's logic will have been vindicated, and pressure on other companies to join will intensify. If the technology proves less revolutionary than hoped, or if its use generates international backlash or domestic political controversy, the calculus may shift. For now, Anthropic remains outside the tent, having made a bet that there is a market for AI systems built with constraints the Pentagon rejected. Whether that bet pays off depends on whether other institutions—governments, corporations, or individuals—value those constraints enough to pay for them.
Citas Notables
The Pentagon framed these partnerships as accelerating the transformation toward establishing U.S. Armed Forces with AI at the center of how they fight.— Pentagon official statement
Anthropic attempted to include clauses vetoing the use of its technology in mass surveillance and lethal autonomous weapons, conditions the Pentagon rejected.— Anthropic's position in contract negotiations
La Conversación del Hearth Otra perspectiva de la historia
Why did the Pentagon move so quickly to sign these seven companies when Anthropic walked away? Why not wait, or negotiate further?
Because the Pentagon sees AI as a race. If American military systems aren't built on American AI, they're vulnerable to dependence on foreign technology or to gaps in capability. Speed matters more than consensus in that logic.
But Anthropic had already been working with them. They had skin in the game. What changed?
Anthropic wanted to add restrictions—no mass surveillance, no autonomous lethal weapons. The Pentagon said no. For Anthropic, those weren't negotiable points; they were foundational to how the company wanted to operate. So they left.
Do you think the Pentagon actually needed Anthropic, or was it easy to replace them?
Probably easier than it should have been. OpenAI, Google, Microsoft—they all have the technical capability. But Anthropic had built something specific around safety and constraint. That's harder to replicate quickly. The Pentagon chose capability and speed over that particular approach.
What about the employees at Google who are objecting? Are they going to slow this down?
Unlikely to stop it, but they're signaling something important: there's real internal disagreement about whether this is the right move. That kind of resistance can create friction, but it rarely kills a Pentagon contract once it's signed.
If Anthropic is right about the risks—surveillance, autonomous weapons—doesn't the Pentagon's move make those risks more likely?
Yes. By excluding the one company that refused to enable those capabilities, the Pentagon has removed a constraint. Whether that matters depends on whether the other companies impose their own limits, or whether they build what they're asked to build without question.
So Anthropic is betting that someone else will value what they're offering?
Exactly. They're betting that ethical constraints on AI will eventually become a selling point, not a liability. It's a long bet, and it only works if the Pentagon's approach creates enough problems that other institutions start demanding alternatives.