A ban or restriction has to come from the federal government (any branch) and has to actually go into effect. If a ban is announced but the effective date is in 2028, I'll extend the close date to it's effective date. I'll buffer to allow for legal challenges within 1 month of effective date (eg: if ban goes into effect Jan 1, but court blocks on Jan 5, I won't count that as a ban).

I will not bet so I can subjectively decide if narrow exceptions count. For the types of exceptions I can imagine:

1. company size limits still count: "companies over X employees can't use Y model" is a ban.

2. narrower usecases do not count: "companies can't use Chinese models for hiring decisions" is not a "full ban on all US companies" but it is a "restriction"

Examples of restrictions that are not bans: limits on models of a given size, some sort of audit requirement, conditions on usecases, conditions on what data it can access, or where it's run (eg: you can only run deepseek through an approved vendor)

For my purposes, "open" implies a model that someone could theoretically run this unrestricted on their own infrastructure.

If the ban/restriction comes from the Chinese side, they won't count.

Market inspired by this article: https://www.cnbc.com/2026/07/08/chinese-ai-models-probe-us-lawmakers.html

Lawmakers probe growing use of Chinese AI models in U.S. companies

An ongoing House Committee investigation is probing the risks involved in the rise of AI built in China.