Anthropic’s Claude AI was used by the US Central Command for intelligence assessments and target identification during strikes on Iran, despite President Trump’s order to halt its use. The Pentagon’s deep integration of the tool made removal impractical, and it was also deployed in January’s operation to capture Nicolás Maduro. A $100 million proposal to develop voice-controlled autonomous drone swarms using Claude was rejected, but the bid highlighted ambitions to turn the AI into a battlefield command system.

The AI’s role in warfare raises ethical concerns. For instance, Lavender, an AI system used by Israel in Gaza, misidentified targets 10% of the time, leading to wrongful deaths. Such errors underscore the dangers of deploying unregulated AI in combat. Critics warn that hallucinations and confabulations in large language models could cause catastrophic mistakes in dynamic combat environments.

Current AI military applications operate in a regulatory vacuum. Article 36 of the Geneva Convention mandates testing new weapons, but AI systems that learn from data evolve too rapidly for traditional frameworks. The US previously concealed drone programs for years, delaying accountability. Experts argue transparency is critical before AI becomes routine in warfare.

Societal debates on AI ethics persist, but battlefield use remains opaque. Without disclosure, mistakes may go unchecked. As AI’s role expands, oversight must evolve to prevent harm. Anthropic’s Claude AI, $100 million proposal, Lavender, and 10% error rate highlight the urgent need for accountability in military AI.