Anthropic warns as Trump deploys AI in warfare, marking dangerous turning point

Anthropic’s Claude model was reportedly employed twice by the Trump administration to facilitate regime‑change operations, first in a covert attempt to capture Venezuelan President Nicolás Maduro and most recently to assist in targeting Iran, according to a report by The Guardian. In the Venezuela case, officials allegedly fed the AI with satellite imagery, diplomatic cables and open‑source intelligence to map security perimeters and identify optimal ingress routes, though the exact algorithms and decision‑making processes remain undisclosed. The Iranian operation, described as a “horribly damaging barrage of missiles,” used Claude to parse massive streams of intelligence, prioritize high‑value targets and run rapid simulations of strike outcomes, a step that the publication says brought the United States “as close to regime change as possible” while leaving the final execution to on‑the‑ground forces.

The use of Claude sparked an immediate clash with Anthropic’s leadership. CEO Dario Amodei publicly refused President Trump’s demand to lift two safety “red lines” that bar the model from mass domestic surveillance and from being integrated into fully autonomous weapons that can select and engage targets without meaningful human oversight. Amodei’s stance prompted Defense Secretary Pete Hegseth to invoke supply‑chain security statutes and label Anthropic a “security risk,” after which the White House issued an executive order banning all federal agencies from employing the company’s AI tools, Reuters reported.

In the wake of Anthropic’s exclusion, OpenAI stepped in to fill the vacuum. The firm secured a comparable defense contract, promising that its deployment architecture would preserve the same ethical guardrails that Anthropic had refused to abandon. According to the Daily AI Rundown newsletter, OpenAI’s agreement includes cloud‑based controls and strict contractual clauses designed to prevent the model from being used for autonomous lethal decision‑making or mass surveillance. However, the rapid transition has ignited a fierce debate within the tech community, with consumer boycotts of ChatGPT and coordinated protests by hundreds of technology workers demanding tighter limits on AI’s role in warfare.

Industry analysts warn that the precedent set by the Trump administration could accelerate a global arms race in AI‑driven weaponry. The Guardian’s piece underscores that the integration of large language models into target‑selection pipelines marks “a dangerous turning point” for militarisation, as the technology can process terabytes of data faster than human analysts and generate actionable strike plans with minimal oversight. If such capabilities become standard practice, the threshold for launching kinetic operations may lower, increasing the risk of miscalculation and civilian casualties.

International bodies are already responding. Reuters cited a statement from the chair of the Geneva talks on lethal autonomous weapons, calling for urgent progress on rules that would restrict AI‑enabled systems from independently selecting and engaging targets. The United Nations’ ongoing deliberations could soon confront the reality that state actors are already field‑testing these technologies, forcing policymakers to grapple with a rapidly narrowing window for effective regulation.

The fallout from the Anthropic episode illustrates a broader tension between national security imperatives and corporate ethical commitments. While the Trump administration argues that AI offers decisive advantages in contested regions, Anthropic’s refusal to compromise its safety safeguards has positioned the company as a focal point for public scrutiny and industry activism. As the United States continues to embed AI deeper into its military apparatus, the balance between innovation, accountability and the preservation of human control remains a contested and increasingly urgent battlefield.