AI-Driven False Alarm Nearly Provokes U.S. Seizure of Chinese Ship

False positives triggering military escalation are not new: the 1964 Gulf of Tonkin misreports drew the U.S. deeper into Vietnam, and the Soviet early-warning computer glitch of 26 Sept 1983 almost launched a nuclear response. What is different in 2026 is the speed and opacity of large-language-model outputs: a junior analyst can now synthesize thousands of signals into a polished, authoritative report in minutes, compressing the OODA (observe–orient–decide–act) loop to near real-time. The Pentagon’s decentralized, “experiment-first” AI rollout reflects a broader 21st-century trend—outsourcing cognition to algorithms without uniform vetting standards—mirroring civilian sectors’ embrace of automated decision-making. Over a 100-year horizon, this incident is a data point in the long shift from human-driven to machine-shaped statecraft; whether it is remembered like Tonkin (as a catalyst for guardrails) or like lesser-known near-misses (quickly forgotten) will hinge on whether governments now impose rigorous validation layers or double down on speed over certainty.