How AI Firms Became Involved in U.S. Military Initiatives

At the beginning of 2024, Anthropic, Google, Meta, and OpenAI stood together against the military deployment of their AI technologies. However, as the year progressed, a significant transformation occurred.

In January, OpenAI discreetly lifted its prohibition on the application of AI for “military and warfare” purposes, subsequently revealing its collaboration on “numerous projects” with the Pentagon. By November, coinciding with Donald Trump’s reelection as US president, Meta declared that the United States and select allies would utilize Llama for defense applications. Just days later, Anthropic announced its own willingness to permit military use of its models, partnering with the defense firm Palantir. As the year concluded, OpenAI unveiled its partnership with the defense startup Anduril. Finally, in February 2025, Google updated its AI principles to accommodate the development and application of weapons and technologies that could cause harm. In just one year, concerns regarding the existential threats posed by AGI virtually vanished, normalizing the military application of AI.

This shift is partly attributed to the substantial investment required to create these models. Research on general-purpose technologies (the other GPTs) frequently underscores the vital role of the defense sector in addressing adoption challenges. Economist David J. Teece noted in 2018, “GPTs evolve more rapidly when there’s a substantial, demanding, and revenue-generating application sector,” referencing the US Defense Department’s expenditures on early transistors and microprocessors. The flexible budget constraints and long-term perspectives of defense contracting, along with often vague success metrics, render the military an attractive client for new technologies. As AI startups seek significant and patient investments, the pivot toward military funding was arguably unavoidable. However, this does not account for the rapidity of the shift or the alignment of all leading American AI research labs in this direction.

Recent years have significantly altered the landscape of capitalist competition—from one driven by neoliberal ideals to one increasingly influenced by geopolitical considerations. To grasp the transition from neoliberalism to geopolitics, one must understand the dynamics between states and their major technology firms. Such state-capitalist relationships have been central to previous forms of imperialism; Lenin famously described the imperialism of his time as a merger of monopoly capital and great powers, a dynamic that continued to exert influence throughout the 20th century. In recent decades, this has manifested as a broad consensus among political and tech elites regarding the role of digital technology in innovation, growth, and state power.

However, this alignment of interests among elite groups has recently fractured. A series of overlapping developments, gaining particular traction in the 2010s, has unraveled this consensus, leaving behind the remnants of potentially new configurations in both the United States and China.

The Silicon Valley Consensus

Until around the mid-2010s, the United States was characterized by what could be termed the Silicon Valley Consensus. There existed a general agreement among both political and tech elites regarding technology’s role in the world, what was necessary for enabling technological advancement, the American values they ostensibly represented, and the prerequisites for capital accumulation in the tech sector. For these elites, globalized communication, capital, data, and technology aligned with their interests.

The Silicon Valley Consensus resonated with both technology and political leaders due to its faith in technology’s ability to forge an American-led world of unbounded commerce and data. Although the tech sector may have started with more utopian ideals than the pragmatic geopolitical realism of the state, both entities recognized the potential for achieving their goals through similar means.

In practice, this resulted in a largely unregulated environment for the tech sector, where regulations were either notably absent or inadvertently supportive. Deregulation, a hallmark of the wider neoliberal era, notably applied to tech firms, which managed to blur existing regulatory lines and “disrupt” established rules. The absence of any significant federal privacy legislation or meaningful action regarding the gig economy’s workforce underscores this readiness to permit digital firms considerable latitude. Under President Bill Clinton, the Framework for Global Electronic Commerce outlined policies that, according to international studies professor Henry Farrell, effectively “discouraged policymakers from attempting to tax or regulate” the digital landscape—instead promoting voluntary, industry-led regulation. The foundational belief here—one that persists today—was that any regulatory intervention would merely obstruct innovation and the growth of US technology and influence.