Sovereign posture and commercial dependency are the same conversation.

I’ve been optimising some of the API calls behind www.aiinasia.com today, and it made me stop and think about the wider impact of sovereign AI and its impact on bigger businesses: ‘if the US restricted access to Chinese models tomorrow, or China restricted something back, what actually breaks in a large company’s stack on Monday morning. I wonder how many SLTs can offer up a clean answer to that question’.

From what I can tell, Silicon Valley’s lobbying against a US ban on Chinese AI models this week is mostly being read as a geopolitics story. I’d read it as a disclosure. When your own suppliers ask their government not to shut off their biggest competitor, they’re telling you their commercial model depends on that competitor staying in the market. That’s the same information a supplier concentration note gives you in an audit pack, just wearing a policy suit.

Most large companies I talk to keep sovereign posture and commercial dependency in separate conversations: one sits with policy and public affairs, the other with procurement and the technology function… but nobody appears to be checking whether the two stories agree - and I’d suggest they probably don’t. The risk register still carries “model provider” as a single line, when what’s behind it is several providers, several jurisdictions, each with a different exposure if the politics moves.

It’s worth double checking which models you actually depend on, under whose law each sits, and what the fallback looks like if one gets cut off without warning.