Clearance wars

The clearance wars are the phase of AI competition in which frontier capability ships not when it is built, but when it is permitted. The term describes a structural shift in what is scarce: once capability is abundant and multiple labs can produce comparable models, access — not intelligence — becomes the contested resource.

The inflection point was GPT-5.6, the first frontier model gated per-customer by US government clearance: capability rolled out tier by tier, customer by customer, with the government rather than the vendor deciding the order. The gate, not the model, became the story. Once that template exists, it generalizes — export controls on coding agents, per-jurisdiction deployment approvals, trust reviews of what runs inside your development environment.

For operators the consequence is a new axis of competitive planning: your AI roadmap is no longer only a function of what exists, but of what you are cleared to use, where, and when. Clearance status becomes an asset with market value — and the vocabulary of AI strategy acquires a word it did not need when capability itself was the frontier.