In what may become one of the most consequential ironies of the emerging artificial intelligence era, the United States military reportedly employed Anthropic’s Claude AI system to assist in planning its recent strikes against Iran’s nuclear facilities — just days after the federal government had moved to restrict the very same company’s technology from certain government applications. The revelation raises profound questions about the relationship between Silicon Valley’s AI firms and the national security apparatus, the enforceability of government technology bans, and the ethical commitments AI companies have made to the public.
According to a report from Engadget, U.S. military officials used Anthropic’s Claude chatbot in the planning stages of Operation Bunker Buster, the American-Israeli joint strikes on Iranian nuclear sites that began in June 2025. The AI was reportedly used to help analyze intelligence, model potential strike scenarios, and assess collateral damage estimates — functions that would place it squarely in the category of military decision-support tools that Anthropic has historically said its technology should not be used for.
A Timeline That Raises Hard Questions
The timeline of events is striking. In late May 2025, the General Services Administration formally removed Anthropic’s products from its approved technology list for federal agencies, citing concerns over data handling practices and contractual disputes. The ban, which applied across civilian agencies, was widely reported as a significant setback for Anthropic’s government business ambitions. Yet within days of that administrative action, elements of the defense and intelligence community were apparently relying on Claude for one of the most sensitive military operations in recent American history.
The apparent contradiction can be partially explained by the fragmented nature of federal technology procurement. The GSA ban applied to civilian agencies and did not necessarily extend to the Department of Defense or intelligence community, which operate under separate acquisition authorities. Military and intelligence agencies often maintain their own technology procurement channels, and classified programs can operate under authorities that supersede standard federal purchasing rules. Still, the optics of the situation are jarring: one arm of the government was distancing itself from Anthropic while another was pulling the company’s technology closer to the center of wartime decision-making.
Anthropic’s Stated Principles Meet Wartime Reality
Anthropic has long positioned itself as the safety-conscious alternative in the AI industry. Founded in 2021 by former OpenAI executives Dario and Daniela Amodei, the company has built its brand around what it calls “responsible scaling” and has published extensive documentation on its efforts to prevent misuse of its AI systems. The company’s acceptable use policy explicitly prohibits the use of Claude for “weapons development” and activities that could cause “mass harm.” Anthropic has also publicly stated that it does not want its technology used for autonomous weapons systems or lethal targeting decisions.
Yet the line between prohibited military applications and permissible intelligence analysis has always been blurry. Using an AI to help plan a military strike does not mean the AI is pulling the trigger, but it does mean the technology is embedded in a kill chain — the sequence of steps that leads from identifying a target to destroying it. Anthropic has not publicly commented on whether it was aware of the military’s use of Claude in the Iran operation, and the company did not respond to multiple requests for comment from reporters at several outlets. The silence itself is notable for a company that has made transparency a core part of its public identity.
The Pentagon’s Growing Appetite for Commercial AI
The military’s use of Claude fits within a broader and accelerating trend of the Pentagon turning to commercial AI products rather than building bespoke systems. The Department of Defense has dramatically increased its spending on artificial intelligence in recent years, with the 2025 defense budget allocating over $2 billion specifically for AI-related programs. Much of that spending has gone to commercial vendors, as the Pentagon has concluded that the private sector is moving faster than government labs in developing capable AI systems.
This shift has created a new kind of tension between technology companies and the military. During the early years of the current AI boom, several major firms — most notably Google, which faced an employee revolt over Project Maven in 2018 — drew bright lines around military work. But those lines have steadily faded. Google quietly resumed defense AI work, Microsoft secured the massive JEDI cloud computing contract (later restructured as JWCC), and Amazon Web Services has built dedicated classified cloud regions for the intelligence community. Anthropic, despite its safety-first branding, has not been immune to these gravitational forces.
How Claude Was Reportedly Used in Operation Planning
Sources familiar with the operation, as cited in reporting by Engadget, indicated that Claude was used in several specific ways during the planning of the Iran strikes. Analysts reportedly fed the system satellite imagery analysis, signals intelligence summaries, and open-source information about Iranian nuclear facilities, asking it to synthesize the data and identify potential vulnerabilities. The AI was also reportedly used to model civilian population patterns near target sites, helping planners estimate the potential for collateral damage under different strike scenarios.
None of these applications involved Claude making autonomous targeting decisions or directly controlling weapons systems. But they do represent a significant use of commercial AI in an active military campaign, and they raise questions about whether the outputs of a commercial chatbot — trained on publicly available data and not specifically validated for military intelligence applications — should be trusted in life-and-death planning scenarios. Military AI systems typically undergo extensive testing and validation before being deployed in operational settings. It is unclear whether Claude was subjected to any such process before being used in the Iran planning.
The Regulatory Gap Between Civilian and Military AI Procurement
The episode exposes a significant gap in how the federal government regulates AI procurement. While civilian agencies are subject to GSA oversight and standardized procurement rules, the defense and intelligence communities operate with considerably more flexibility. Executive orders on AI safety issued by both the Biden and Trump administrations have focused primarily on civilian applications, leaving military AI use governed largely by internal Pentagon policies and the judgment of individual commanders and program managers.
Members of Congress have begun to take notice. Senator Edward Markey of Massachusetts, a longtime advocate for AI regulation, issued a statement calling the reports “deeply troubling” and demanding a full accounting of which commercial AI systems are being used in military operations. “The American people deserve to know whether the companies building the most powerful AI systems in the world are complicit in military operations, and whether those companies’ safety commitments are worth the paper they’re printed on,” Markey said. Representative Ted Lieu of California, who has been one of the most active voices on AI policy in the House, called for hearings on the matter.
Anthropic’s Business Ambitions and the Defense Dollar
For Anthropic, the situation creates an acute strategic dilemma. The company has raised over $15 billion in funding and is reportedly seeking a valuation north of $60 billion in its next round. Its investors include Google, Salesforce, and a roster of major venture capital firms. Growth at that scale requires enormous revenue, and government contracts — particularly defense and intelligence contracts — represent some of the largest and most reliable sources of technology spending in the world.
Walking away from military business would be consistent with Anthropic’s stated principles but potentially devastating to its financial trajectory. Embracing it would alienate a significant portion of the AI safety community that has supported the company precisely because of its ethical commitments. The company appears to be threading this needle by maintaining its public policy positions while not actively preventing the government from using its commercially available products — a posture that critics have described as willful blindness.
What This Means for the Future of AI in Warfare
The use of Claude in the Iran strike planning is unlikely to be an isolated incident. As AI systems become more capable and more deeply integrated into military workflows, the boundary between commercial technology and military technology will continue to erode. The question is not whether AI will be used in warfare — it already is, and has been for years in various forms — but whether the companies building these systems will have any meaningful say in how they are used, and whether the public will have any visibility into those decisions.
The Iran episode also underscores the speed at which AI capabilities are being absorbed by national security institutions. Just three years ago, large language models were primarily novelties, capable of generating text but not trusted for serious analytical work. Today, they are apparently being used to help plan strikes on a nation’s nuclear infrastructure. The pace of adoption has outstripped the development of governance frameworks, oversight mechanisms, and ethical guidelines — a gap that policymakers, technologists, and military leaders will need to address with urgency before the next operation is planned with an AI assistant at the table.
For Anthropic, the path forward will require more than silence. The company that built its reputation on being the responsible actor in AI will eventually need to explain how its technology ended up in the middle of a war — and whether it intends to do anything about it.
WebProNews is an iEntry Publication