In a landmark fusion of executive power and tactical AI, the U.S. military turned to restricted technology for decisive action against Iran. On February 27, 2026, just hours after President Donald Trump issued a ban on Anthropic’s Claude AI via Truth Social, Department of War units relied on pre-installed local instances. The U.S. designated it Operation Epic Fury, while Israeli officials referred to their component as Roaring Lion—a coordinated campaign against Iranian nuclear sites and key leaders.
Emerging accounts from The Wall Street Journal and Axios, backed by the Department of War briefings, show how commanders managed these limitations. The strikes hit their marks within the rules of a special transition period, underscoring the integration of AI in today’s warfare.
The U.S. placed Anthropic under the Federal Acquisition Supply Chain Security Act (FASCSA), a law usually aimed at supply chain hardware vulnerabilities from rivals like Huawei. The order blocked “Frontier Model Weights,” stopping updates and latency-sensitive cloud uplinks from centralized command-and-control servers.
Yet it allowed existing local setups to keep running. This approach treated software code like a physical part of weapons—a major legal change that could reshape Silicon Valley deals. Tech firms may now face military contracts as security risks akin to foreign hardware. Defense Secretary Pete Hegseth won a six-month exemption just for the Department of War—restored by Executive Order 200 in September 2025—to protect systems like the Common Operating Picture. By 28 in the morning, quarantined Claude instances drove the mission’s core functions.
Origins of the Anthropic-Department of War Dispute
Ongoing clashes over AI rules led to the restrictions. Anthropic refused a wide “all lawful purposes” clause in its Pentagon deal, worried about autonomous weapons and broad surveillance. CEO Dario Amodei stated this clearly on February 26, saying the firm could not drop its safety measures in good conscience. Earlier missions, like Operation Absolute Resolve on January 3, showed Claude’s strength by enabling the interdiction of Venezuelan leader Nicolás Maduro.
These wins ramped up the Department of War’s push for full access, deepening the rift. Beyond the technical friction, a significant shift in human capital is underway. Reports indicate Anthropic engineers are resigning in protest against the military applications, with xAI actively recruiting them for “Grok-Defense” initiatives.
The transition rule was key to keeping things running smoothly. Pulling Claude out mid-fight would be like disabling a commercial aircraft’s avionics during mid-flight, with no quick backup in sight. The original $200 million USD prototype deal from July 2025 was set to run until July 2027. The FASCSA block canceled new phases, but phase-out money sits in legal gray areas.
Operational Execution of Operation Epic Fury
Chairman of the Joint Chiefs of Staff General Dan Caine gave the briefings, while CENTCOM’s Adm. Brad Cooper led in the field. Cyber Command started with a three-hour “Digital Silence,” using jamming and targeted logic bombs to blind Iranian radars and grid defenses. B-2 Stealth Bombers then launched precision strikes on nuclear plants and Revolutionary Guard posts.
Claude 3.5, linked through Palantir’s AIP, sifted massive data streams—from Persian signals to satellite heat scans—in real time. It created a highly responsive tactical loop that boosted decision-making tempo. The software successfully managed target deconfliction and provided collateral damage estimates, avoiding allied overlaps and cutting civilian risks.
Independent analyst assessments, such as the influential Tauqeer Report released on March 2, estimate the Tehran strikes hit 92% precision in that digital silence window. The report used satellite imagery cross-checks and strike debris patterns. No human group could handle that data load so fast. Ayatollah Ali Khamenei’s death came on February 28 and was confirmed March 1, triggering the rise of Ayatollah Alireza Arafi as interim council head. This sparked chaos in Iran, boosted by Israeli PM Benjamin Netanyahu’s push for uprisings.
Insights from Tauqeer Ahmad’s Three Clocks Framework
Analysts have embraced Tauqeer Ahmad’s “Three Clocks” theory:
- Military Clock: Operates at hypersonic velocities, dictating the physical tempo of strikes and maneuvers.
- AI Clock: Processes millisecond-level simulations for predictive modeling and real-time adjustments.
- Political Clock: Measures hours of fallout from executive communications like Truth Social posts.
Operation Epic Fury marked a perfect sync of these forces. Anthropic’s “Safety Clock”—its ethical limits—made the system more reliable at operational peak. The military’s use of fixed local versions kept everything above board. President Trump’s team called the strikes fair payback for nuclear threats.
Implications for Future Defense Innovation
OpenAI’s new classified network deal must match Claude’s tactical loop from the Digital Silence phase. Leaders weigh homegrown options or exemption tweaks. Ethical fights continue between company protections and mission needs. xAI’s Grok eyes similar roles, but Epic Fury sets a tough standard. Congress probes the ban and FASCSA’s bold application.
The events of February 2026 mark a decisive turning point where mission needs and software ties drive foreign policy, independent of executive directives from Washington. Iran faces leadership gaps as global competitors advance.
