This Strategic Briefing has been developed in the immediate aftermath of the public confrontation between Anthropic and the U.S. Department of Defence over the use of frontier AI systems for all lawful purposes. It is the first iteration of a focused body of work on AI red lines: the points at which vendor ethics, state authority and enterprise risk collide.

This briefing draws on interdisciplinary work across law, security studies, technology policy and enterprise risk management, as well as ongoing conversations with practitioners in government, defence, large enterprises, model labs and civil society. Those dialogues, and the shockwaves from the Anthropic–Pentagon standoff, revealed a growing gap between how AI is procured, how it is actually governed in practice, and where different actors believe the moral and legal off‑switch should sit.