NextFin News - Anthropic CEO Dario Amodei escalated a high-stakes confrontation with the Pentagon on Thursday, accusing rival OpenAI of predatory behavior and "safety theater" as the U.S. military shifts its multi-billion dollar artificial intelligence strategy toward more permissive models. The dispute, which reached a fever pitch on March 5, centers on Amodei’s claim that OpenAI CEO Sam Altman worked behind the scenes to have Anthropic designated a "supply chain risk" to clear the path for a massive, exclusive contract with the Department of Defense.
The rift between the two AI giants and the Trump administration has transformed from a technical debate over guardrails into a raw political brawl. According to reports from the Financial Times and Bloomberg, Amodei informed staff and investors that the Pentagon’s recent decision to freeze Anthropic out was "retaliatory and punitive," stemming from the company’s refusal to provide what he termed "dictator-style praise" to U.S. President Trump. This defiance led Defense Secretary Pete Hegseth to label Anthropic a supply chain risk—a designation usually reserved for adversarial foreign entities like Huawei—effectively barring military contractors from using Anthropic’s Claude models.
OpenAI moved with clinical speed to fill the vacuum. As the Trump administration severed ties with Anthropic, Altman secured a deal that Amodei now characterizes as "80% safety theater." The Anthropic chief alleges that while Altman publicly advocated for industry-wide standards, he was privately advising the Department of War on how to replace Anthropic the moment the supply chain risk designation was finalized. This maneuver has effectively handed OpenAI a near-monopoly on the next generation of military decision-support systems, a market estimated to be worth tens of billions over the next decade.
The financial stakes are matched by a fundamental disagreement over the "red lines" of autonomous warfare. Anthropic has historically insisted on rigid safety protocols that prevent its AI from being used in lethal targeting or strategic nuclear planning. The Pentagon, under the direction of U.S. President Trump, has signaled a preference for "unfettered" AI capabilities, viewing Anthropic’s caution as a competitive disadvantage against China. By labeling these safeguards as a risk to national security, the administration has signaled that "safety" is now viewed through the lens of speed and dominance rather than ethical restraint.
Despite the vitriol, a total divorce remains unlikely. The Pentagon already relies heavily on Anthropic’s architecture for non-lethal logistics and data processing, and an immediate transition to OpenAI’s systems would be technically disruptive. Under-secretary of defense Emil Michael has reportedly resumed "last-ditch" talks with Amodei to find a compromise, even as he publicly dismissed the CEO as having a "God complex." For Anthropic, the path forward is narrow: it must either dilute its safety principles to satisfy the administration’s "America First" AI doctrine or face a permanent exile from the world’s largest defense budget.
The outcome of this standoff will dictate the ethical framework of the first AI-driven arms race. If OpenAI’s more permissive approach becomes the default for the U.S. military, the industry-wide "safety" movement that birthed Anthropic may find itself relegated to the civilian sector, leaving the most powerful tools of the state governed by a different set of rules. Amodei’s decision to go public suggests he believes the court of public opinion—or perhaps a legal challenge to the "supply chain risk" label—is his last remaining leverage against a consolidated front of political and corporate rivals.
Explore more exclusive insights at nextfin.ai.
