For years, the leading artificial intelligence companies in San Francisco positioned themselves as guardians of a technology too powerful and too unpredictable to hand over to the military. OpenAI’s founding charter spoke of ensuring AI benefits “all of humanity.” Anthropic built its brand around “AI safety.” The implicit — and sometimes explicit — message was that the Pentagon’s warfighting mission was incompatible with their corporate values.
That era is over. In a dramatic and accelerating shift, both OpenAI and Anthropic have moved to embrace contracts with the U.S. Department of Defense, a transformation driven by a volatile mix of geopolitical competition with China, enormous government spending opportunities, and direct political pressure from the Trump administration. As reported by TechCrunch, the two most prominent AI labs have now formally signaled their willingness to work with the U.S. government on defense and national security applications, marking a turning point for the industry and raising profound questions about the future relationship between Silicon Valley and the American military establishment.
From Red Lines to Green Lights: The Policy Reversals
OpenAI’s pivot has been the most visible. The company, led by CEO Sam Altman, quietly removed language from its usage policies in early 2024 that had previously prohibited military and warfare applications of its technology. That policy change, which drew immediate scrutiny from researchers and civil liberties organizations, was the first clear signal that the company was preparing to pursue government defense work. By mid-2024, OpenAI had begun engaging with defense and intelligence agencies, and by early 2025, the company was actively bidding on Pentagon contracts.
Anthropic’s shift has been more measured but no less significant. The company, founded by former OpenAI researchers Dario and Daniela Amodei, had long maintained that its focus on AI safety made it a natural fit for ensuring that powerful AI systems are deployed responsibly — including in government settings. But the company’s willingness to engage directly with Department of Defense programs represents a departure from its earlier posture, which emphasized civilian and commercial applications. According to TechCrunch, both companies have now moved beyond exploratory conversations and into active collaboration with defense officials.
Pete Hegseth and the White House Pressure Campaign
The political context for these moves cannot be overstated. Defense Secretary Pete Hegseth has made the integration of AI into military operations a centerpiece of his agenda at the Pentagon. Hegseth, a former Fox News host who took the helm at the Department of Defense in early 2025, has publicly called on American technology companies to support the nation’s defense posture, framing the issue as one of patriotic duty and competitive necessity. His argument is straightforward: if the most capable AI systems in the world are built by American companies, those companies have an obligation to ensure that the U.S. military — not just China’s People’s Liberation Army — benefits from the technology.
Hegseth’s approach has combined public exhortation with behind-the-scenes arm-twisting. Administration officials have made clear that companies seeking favorable regulatory treatment on AI — including lighter-touch oversight and fewer export restrictions on chips — will be expected to demonstrate their willingness to support national security objectives. The message has been received. As TechCrunch detailed, the convergence of political pressure and commercial incentive has created conditions under which even the most safety-conscious AI companies have found it difficult to remain on the sidelines.
The China Factor: A Race That Tolerates No Bystanders
The single most powerful argument driving AI companies toward defense work is the intensifying technology competition with China. Beijing has made no secret of its ambitions to achieve global leadership in artificial intelligence by 2030, and the Chinese military has been aggressively integrating AI into surveillance, autonomous weapons research, logistics, and command-and-control systems. The emergence of capable Chinese AI models — including those from companies like DeepSeek, Baidu, and Alibaba — has sharpened the sense of urgency in Washington.
For OpenAI and Anthropic, the China argument provides both strategic rationale and political cover. Company executives have increasingly adopted the framing that American AI leadership is not guaranteed and that refusing to work with the U.S. military does not prevent adversaries from deploying similar technology — it merely ensures that the United States falls behind. Dario Amodei has spoken publicly about the national security implications of AI, arguing that responsible engagement with government is preferable to ceding the field to less safety-conscious actors, whether foreign governments or domestic competitors with fewer scruples about deployment.
What the Contracts Actually Look Like
The specific nature of the defense work being pursued by OpenAI and Anthropic remains partially obscured by classification and corporate confidentiality. However, reporting from multiple outlets suggests that the initial contracts focus on areas that the companies can frame as defensive or administrative rather than directly lethal. These include intelligence analysis and summarization, where large language models can process vast quantities of intercepted communications, satellite imagery reports, and open-source intelligence far faster than human analysts. Logistics and supply chain optimization is another area of focus, along with cybersecurity applications, where AI models can detect and respond to network intrusions in real time.
Pentagon officials have also expressed interest in using commercial AI models for war-gaming and scenario planning, allowing military planners to simulate complex geopolitical situations and test potential responses. None of these applications involve AI systems directly making kill decisions — a line that both companies have publicly said they will not cross. But critics note that the boundary between “back-office” AI and battlefield AI is far more porous than corporate talking points suggest. An AI system that identifies a target in intelligence data is, functionally, part of the kill chain, even if a human being makes the final decision to strike.
The Internal Tensions and Talent Exodus
The pivot to defense work has not come without internal cost. Both OpenAI and Anthropic have experienced dissent from employees who joined these companies specifically because of their stated commitments to safety and civilian-oriented research. At OpenAI, the policy change in 2024 contributed to a wave of departures among researchers who felt the company was abandoning its founding principles. Several prominent safety researchers left for academic positions or smaller labs, citing concerns about the direction of the organization.
Anthropic has faced similar, if somewhat quieter, tensions. The company’s culture has long emphasized thoughtful deliberation about the societal impacts of AI, and some employees have expressed discomfort with the speed at which defense partnerships have materialized. However, Anthropic’s leadership has argued that engagement with the military is itself a form of safety work — that it is better to have safety-focused organizations shaping how AI is deployed in national security contexts than to leave that work to companies with less rigorous safety practices. This argument has persuaded some skeptics internally, but it has not eliminated the friction.
The Broader Industry Realignment
OpenAI and Anthropic are not operating in isolation. Their moves reflect a broader realignment across the technology industry. Google, which faced an employee revolt in 2018 over Project Maven — a Pentagon program that used AI to analyze drone footage — has quietly resumed defense work through its Google Public Sector division. Microsoft, which invested billions in OpenAI, has long maintained an active defense contracting business through its Azure Government cloud platform. Palantir, Anduril, and Shield AI have built their entire businesses around defense applications of advanced technology.
What makes the OpenAI and Anthropic moves distinctive is the symbolic weight they carry. These were the companies that were supposed to be different — the ones that would prove that it was possible to build transformative AI technology while maintaining ethical boundaries that excluded military applications. Their entry into the defense market signals that the industry’s brief experiment with principled abstention from military work has effectively ended. The gravitational pull of government contracts, political pressure, and great-power competition has proven stronger than the idealism that characterized the AI safety movement’s early years.
What Comes Next for AI and the Pentagon
The implications of this shift extend well beyond the balance sheets of two San Francisco companies. The integration of frontier AI models into U.S. military operations will raise urgent questions about accountability, transparency, and the laws of armed conflict. International humanitarian law requires that weapons systems allow for meaningful human control over targeting decisions. As AI systems become more deeply embedded in military command structures, the line between human decision-making and machine recommendation will become increasingly difficult to discern.
There are also questions about what happens when these companies’ commercial interests conflict with national security imperatives. OpenAI and Anthropic operate globally, with users and customers in dozens of countries. Defense work could complicate their international operations, particularly in Europe, where regulators have taken a more cautious approach to AI governance. And the precedent being set now — that the most capable AI systems should be made available to the military — will be difficult to reverse if political conditions change.
For the moment, however, the direction of travel is clear. The AI industry’s most prominent companies have made their choice, and the Pentagon is the beneficiary. Whether that choice ultimately serves the interests of national security, corporate shareholders, or the broader public remains an open and deeply consequential question.
WebProNews is an iEntry Publication