Sometimes a harmless post on X turns into a subpoena. That’s more or less what happened to DoorDash. On July 31, two U.S. House committees asked the delivery company to hand over documents about its use of Chinese AI models.
What happened
The chairmen of the Committee on Homeland Security and the Select Committee on the Chinese Communist Party are running a joint investigation. Their focus: the security risks when U.S. companies deploy AI models built in China. DoorDash is now expected to share “information and documents” on how it evaluates and deploys such systems.
The whole thing was set off by DoorDash co-founder Andy Fang. In a post on X, he described how the company offloads lower-level AI work to Moonshot AI’s Chinese model Kimi K2.6. DoorDash’s AI lab went further: it said Kimi K2.6 and — note this — Anthropic’s Fable 5 had vastly outperformed the other Anthropic models it had been using, and at a lower cost.
Why it’s touchy
Here’s what makes lawmakers nervous: open-weight Chinese models keep getting better and cheaper. For a company processing millions of requests a day, that’s a very concrete argument. The committees themselves admit firms have good reasons to reach for these models — strong performance, low cost. But, they add, that doesn’t erase the national security concerns or the need for safeguards.
DoorDash is just the latest case in a wider campaign. Back in early July, the same committees asked more broadly about Chinese models in U.S. companies. Now it’s getting specific — with a name attached and a public X post as evidence.
My take
What fascinates me about this story is the tension. On one side, pure engineering logic: use the model that does the job best and cheapest. Andy Fang simply said out loud what plenty of companies do quietly. On the other side, a geopolitical reality where a model’s origin suddenly matters more than its benchmark score.
The cameo by Anthropic is the interesting part. Fable 5 gets praised in the same breath as Kimi — a U.S. model that apparently wins the internal comparison. For Anthropic, that’s free advertising at exactly the right moment. Because if Chinese models become politically radioactive, the Western frontier labs hold the stronger hand. The real question stays open, though: can “best model” and “safest choice” be kept apart forever? Companies like DoorDash will need an answer soon.
Sources: