Tech Industry's Military Contracts Controversy

The collision between national security demands and ethical red lines has fractured the AI industry — and the fallout is just beginning.

On March 7, 2026, Caitlin Kalinowski, the head of robotics at OpenAI, announced she was walking away. In a public statement, she said that "surveillance of Americans without judicial oversight and lethal autonomy without human authorization are lines that deserved more deliberation than they got" [1]. Her departure was not an isolated act of conscience. It was the latest tremor in an escalating confrontation between Silicon Valley and the Pentagon that has split the AI industry down the middle, pitting national security imperatives against the ethical commitments that many technologists consider foundational to their work.

The crisis has its roots in a U.S. military raid in Venezuela, a blacklisting unprecedented in American corporate history, and a rushed defense contract that even its own architect admitted was "opportunistic and sloppy" [2]. Together, they have forced a reckoning: in an era of AI-powered warfare, who gets to draw the red lines?

From Venezuela to the Pentagon: How the Crisis Began

The chain of events that led to Kalinowski's resignation started thousands of miles from San Francisco. On January 3, 2026, U.S. special operations forces captured Venezuelan President Nicolás Maduro in Caracas. Eighty-three people were killed, including 47 Venezuelan soldiers [3]. In the weeks that followed, media reports revealed that Anthropic's Claude AI had been used during the operation through the company's partnership with defense contractor Palantir [4].

Anthropic had partnered with Palantir to allow Claude to be used in government settings — specifically "to support government operations such as processing vast amounts of complex data rapidly" and "helping U.S. officials to make more informed decisions in time-sensitive situations" [3]. But when Anthropic executives learned their technology may have been deployed in an active military raid, the company contacted Palantir to investigate [4].

What followed was a policy dispute that escalated into a full-blown geopolitical confrontation. The Department of Defense — now rebranded as the Department of War under the Trump administration — demanded that Anthropic remove all safeguards and restrictions on military use of its Claude models [5]. Anthropic refused to budge on two non-negotiable positions: no mass domestic surveillance of Americans, and no fully autonomous weapons without human oversight [6].

The Blacklisting: A Penalty Reserved for Adversaries

On February 28, 2026, Defense Secretary Pete Hegseth designated Anthropic a "supply chain risk to national security" — a classification typically reserved for companies from adversarial nations like China [7]. Anthropic became the first American company ever to be publicly labeled with the designation, joining the ranks of entities like Huawei [7]. President Trump directed federal agencies to stop using Anthropic's technology after a six-month transition period [8].

The consequences rippled across the defense technology ecosystem. Military contractors scrambled to remove Anthropic's Claude from their systems to preserve their Pentagon business [9]. The move set off what PitchBook described as a "mad dash" among defense tech startups to pivot away from Anthropic, threatening partnerships that had been years in the making [10].

Anthropic responded by announcing it would challenge the designation in court, rejecting Hegseth's claim that the label would bar military contractors from working with the company [7]. In a public statement, the company wrote: "We believe the tools we build should help defend democracies — but not at the cost of the values that make democracies worth defending" [6].

OpenAI Steps In — And Stumbles

Hours after the Anthropic blacklisting, OpenAI struck a deal with the Pentagon to deploy its AI systems inside classified military networks [2]. The timing was immediately controversial. OpenAI CEO Sam Altman acknowledged the optics, telling reporters the agreement was hastily reached and that "it just looked opportunistic and sloppy" [2].

The original agreement lacked explicit prohibitions on domestic surveillance — a glaring omission that critics were quick to highlight [11]. Under pressure, OpenAI amended the contract within days, adding language stating that its AI models "shall not be intentionally used for domestic surveillance of U.S. persons and nationals" [11]. The Pentagon also affirmed that OpenAI's services would not be used by intelligence agencies such as the NSA [12].

But for many inside OpenAI, the damage was already done. Kalinowski, in a follow-up statement clarifying her resignation, framed her concern not as opposition to all military use of AI, but as a governance failure: "This was a governance concern first and foremost" — the announcement had been "rushed without guardrails defined" [1].

The amended contract now includes "red lines" prohibiting domestic mass surveillance, autonomous weapons without human control, and "high-stakes" AI decisions without human approval [12]. Yet critics, including MIT Technology Review, argued that OpenAI's "compromise" represented exactly the kind of weakened safeguards that Anthropic had feared would become the new standard [13].

The Worker Revolt: "We Will Not Be Divided"

Source: GDELT Project

Data as of Mar 8, 2026CSV

The Anthropic blacklisting and OpenAI's Pentagon deal catalyzed the most significant employee revolt in the tech industry since Google's Project Maven controversy in 2018. Almost 900 employees from Google and OpenAI signed an open letter titled "We Will Not Be Divided," calling for clear limits on military applications of AI [14]. The letter grew from a few hundred names on a Friday to nearly 900 by Monday — with approximately 100 signatories from OpenAI and close to 800 from Google [14].

The letter called on Google and OpenAI leadership to "put aside their differences and stand together" against the Pentagon's demands [15]. Its specific demands included quarterly transparency reports on all government AI deployments, third-party ethics audits, and employee seats on internal review boards that evaluate military contracts [16].

The backlash intensified further after U.S. strikes on Iran over the same weekend, raising immediate questions about whether AI tools from these companies were already being used in active combat operations [17]. Hundreds of additional workers from companies including Salesforce, Databricks, IBM, and Cursor signed a separate open letter urging the Defense Department to withdraw its supply chain risk designation of Anthropic [15].

Google finds itself in a particularly complex position. The company abandoned its Project Maven drone AI contract in 2018 after thousands of employees signed a petition and several resigned [18]. Now, Google is reportedly in talks with the Pentagon about bringing its Gemini AI model onto a classified system — reviving the same internal fight that convulsed the company eight years ago [16].

The Defense Tech Gold Rush

The ethical crisis unfolds against the backdrop of an unprecedented surge of money flowing into military AI. The Pentagon's fiscal year 2026 budget request includes $66 billion for IT spending — a $1.8 billion increase from 2025 — with $9.8 billion directed specifically toward autonomous and unmanned systems [19]. The AI-in-military market is valued at $22.4 billion in 2026 and projected to reach $101 billion by 2034 [20].

Source: DOD Budget Request / Washington Technology / PR Newswire

Data as of Mar 8, 2026CSV

All military branches are increasing AI investment. The Navy allocated $308 million for AI in fiscal 2026, a 22.7 percent increase from 2025. The Air Force committed $415 million, up 21.7 percent [20]. Meanwhile, venture capital firms poured more than $28 billion into defense technology in just the first half of 2025 — already surpassing most recent full-year totals [21].

Companies like Palantir and Anduril have positioned themselves at the center of this boom. Palantir secured an Army enterprise agreement worth up to $10 billion [22], while Anduril, valued at $30.5 billion, is building a $1 billion manufacturing facility in Ohio for weapons systems including AI-equipped drones [23]. The two companies, along with SpaceX, OpenAI, and Scale AI, have formed a consortium to jointly bid for military contracts, representing a new generation of "neoprimes" challenging legacy defense contractors like Lockheed Martin and Boeing [23].

Yet even amid this growth, the Anthropic saga has introduced new uncertainty. A senior Pentagon official told Fortune about a "whoa moment" when defense leaders realized how deeply embedded Anthropic's technology had become in military operations — and how disruptive it would be to lose access [24].

The Broader Stakes: Who Decides?

The confrontation raises questions that extend far beyond any single contract. At its core, it is a dispute over governance: when AI systems are deployed in matters of life and death, who has the authority to set boundaries?

The Pentagon's position is clear. A senior U.S. official warned in early March that "AI contract restrictions could threaten military missions" [25]. The Defense Department argues it needs unrestricted access to the best available AI tools to maintain strategic advantage, particularly as China and Russia develop their own military AI capabilities.

Anthropic has staked out an equally firm position, arguing that certain uses of AI — mass surveillance of citizens, weapons that fire without human authorization — represent absolute limits regardless of the customer. The company's stance echoes a broader principle in the responsible AI community: that the companies building the most powerful AI systems have a responsibility to ensure they are not used to undermine democratic values.

OpenAI, meanwhile, has attempted to occupy a middle ground that has satisfied neither side. Its amended contract asserts red lines while acknowledging — as Altman himself stated — that the company doesn't "get to make operational decisions" about how the military uses its technology [26]. Critics argue this renders the safeguards largely aspirational.

The Electronic Frontier Foundation has weighed in, warning that "tech companies shouldn't be bullied into doing surveillance" and arguing that the Anthropic blacklisting sets a dangerous precedent in which the government punishes companies for maintaining ethical standards [27].

A Precedent With No Parallel

The designation of Anthropic as a supply chain risk stands as the most aggressive action the U.S. government has ever taken against a domestic technology company over an ethical dispute. It has transformed a business negotiation into a constitutional question about the relationship between the tech industry and the national security state.

For the nearly 900 employees who signed the "We Will Not Be Divided" letter, for Caitlin Kalinowski who chose principle over position, and for the defense tech startups now scrambling to realign their businesses, the message is clear: the era in which AI companies could remain agnostic about military applications is over. Every company building frontier AI must now answer a question that, until recently, seemed hypothetical: where do you draw the line?

The answer, as this crisis has demonstrated, carries consequences that reach far beyond Silicon Valley.