Export controls: the licence policy is the policy

An export control is a licensing requirement, not a prohibition. The underlying statute lets the Commerce Department control dual-use items - goods, software and technology with both civilian and military application - for national security and foreign policy reasons, and the mechanism is that certain transactions require permission before they happen. Whether permission is granted is a separate question from whether it is required, and that separation is where nearly all the movement happens.

The Entity List is the designation layer. Named foreign companies and institutions are added to it, and a designation attaches a licence requirement with a stated review policy - typically a presumption of denial. Additions come in batches: forty-two Chinese entities in March 2025, twenty-three more that September, and a further twenty-eight including major domestic server and processor makers, each batch published with its justification.

The foreign direct product rule is the layer that gives all of this global reach, and it is the part most often missed. It captures goods manufactured entirely outside the United States if they were made using American technology or software. Because the tooling and design software underlying advanced chipmaking is overwhelmingly American, the rule reaches fabs in third countries that never touch US soil - which is exactly what stops an allied manufacturer from simply backfilling restricted supply.

The stated purpose is to slow the acquisition of computing capability with military application, and the design follows from that: restrict the most advanced processors, restrict the equipment that makes them, and restrict the ability of US persons to support restricted activity. Those three legs plus the extraterritorial rule are the whole regime, and they have remained in place across changes of administration even as the licence policy has moved considerably.

The practical use is as a dial rather than a switch. In December 2025 an announcement allowed certain high-end accelerators to ship to approved customers in China, and the agency began reviewing applications for those products case by case, subject to security requirements. No rule was repealed. The items remained controlled and the list remained in force; what changed was the answer applications receive - which is the entire commercial question.

For markets, that distinction is the whole trade. Contracts written on whether an item is 'banned' or 'allowed' are asking about a state that rarely changes cleanly. Contracts written on whether a specific entity is designated, or whether a specific rule takes effect by a date, resolve against the Federal Register.

Everything material appears in the Federal Register. Entity List additions and removals, rule changes and effective dates are published as rules with dates and justifications, and the agency issues press releases alongside the significant ones. For a market on whether a company is designated by a date, this is a clean, dated, archived source - the same quality as the sanctions instruments described in the Russia and Venezuela entries.

Licence review policy is published less formally and matters more. It appears in agency statements, in guidance, and sometimes only in the pattern of what applicants report receiving. That asymmetry is worth internalising: the most commercially significant variable in the regime is also the least cleanly documented, which is why so much of the trading around it happens on company disclosures rather than on government publications.

The affected companies are the third source and often the fastest. A restriction that bites appears in guidance, in revenue disclosures broken out by region, and in the specific language public companies use about licence applications in their filings. Those documents are audited, dated, and legally consequential to get wrong - which makes them a better read on what is actually happening than most commentary.