The Intercept reports AI hallucinations in US military weapons tests

The Intercept reported on September 28, 2026, that U.S. military weapons platforms utilizing artificial intelligence demonstrated dangerous hallucinations and susceptibility to adversarial manipulation during testing. According to the article, these technical failures raise significant concerns about deploying generative AI models in lethal autonomous systems amid ongoing strategic competition with China. Critics argue that relying on probabilistic models for targeting or threat assessment introduces unacceptable risks of catastrophic error in high-stakes combat environments. The report alleges that current validation protocols are insufficient to prevent false inputs from triggering unintended kinetic responses. Defense officials have not publicly confirmed the specific test failures cited in the investigation. This disclosure intensifies existing debates regarding the safety standards required for AI integration in national security infrastructure. The findings suggest a gap between commercial AI capabilities and the deterministic reliability demanded by military doctrine.