Anthropic Shows How AI Is Shaping Modern Warfare and What Comes Next

The United States’ Maven Smart System, which integrates Anthropic’s Claude large‑language model (LLM) for image processing and target prioritisation, has been a cornerstone of the recent US‑Israel‑Iran offensive, according to a March 5 2026 report in Nature (Jones). The system, funded by a $200 million contract signed in 2024, feeds real‑time battlefield data to Claude, allowing operators to generate rapid threat assessments and suggest strike orders. While the exact algorithms remain classified, the Nature article notes that Maven “speeds up attack capabilities by suggesting and prioritising targets” and has been deployed in prior conflicts as well as the current strikes on Iran (Horowitz). This operational reliance on LLM‑driven decision support marks a shift from traditional rule‑based automation to generative AI that can parse unstructured visual feeds and produce actionable intelligence on the fly.

The deployment of AI‑guided missiles in the Middle East, highlighted by Nature, underscores how generative models are moving beyond advisory roles into direct weapon control. Missiles equipped with on‑board AI can adjust flight paths in response to sensor inputs, a capability that analysts such as Craig Jones of Newcastle University warn could accelerate the proliferation of lethal autonomous weapons (Jones). However, the same report cautions that current LLM‑powered systems lack the reliability required to meet international humanitarian law standards, which demand unequivocal discrimination between combatants and civilians. Horowitz emphasizes that “LLM‑powered, fully autonomous weapons without human oversight are not currently reliable and do not comply with international laws,” suggesting a regulatory gap between rapid technological adoption and legal frameworks.

Parallel to battlefield integration, the ethical debate is intensifying in diplomatic circles. This week, legal scholars and policymakers convened in Geneva to discuss lethal autonomous weapon systems (LAWS) and the procurement of AI for military use, a meeting that reflects long‑standing attempts to forge an international treaty on AI in warfare (Nature). Michael Horowitz of the University of Pennsylvania warns that “the current failure to regulate AI warfare, or to pause its usage until there is some agreement on lawful usage, seems to suggest potential proliferation of AI warfare is imminent.” The urgency is amplified by the United States’ recent decision to sideline a major AI supplier over ethical concerns just one day before the offensive began, indicating internal friction over the balance between operational advantage and moral responsibility (Nature).

Despite the controversy, proponents argue that AI could reduce civilian casualties by improving targeting precision. The Nature piece cites ongoing conflicts in Ukraine and Gaza where AI assists in target identification and drone navigation, yet acknowledges that “there is no evidence that AI lowers civilian deaths or wrongful targeting decisions and it may be that the opposite is true” (Jones). This ambivalence reflects the nascent state of AI‑enabled weaponry: while generative models can process vast data streams faster than human analysts, their probabilistic nature can produce false positives, especially in complex urban environments where distinguishing combatants from non‑combatants is notoriously difficult.

Looking ahead, Anthropic’s roadmap suggests an expansion of Claude’s capabilities beyond battlefield logistics into broader autonomous systems. Recent coverage from VentureBeat notes that Anthropic is rolling out “Claude Code” and “Claude Cowork,” tools aimed at transforming programming and enterprise workflows (VentureBeat). Although these products target civilian markets, the underlying LLM architecture is the same that powers Maven, raising concerns that the same technology could be repurposed for more sophisticated autonomous weapons. As the gap between AI development and international regulation widens, the next phase of AI‑driven warfare may hinge on whether policymakers can impose meaningful oversight before the technology becomes entrenched in national defense strategies.