The White House worries China got into Anthropic's AI

A new Semafor report says the White House is moving to restrict exports of Anthropic's Mythos model partly because Chinese-linked groups may have already accessed it. If true, that's more than a data leak — it's a potential back door. The government could use distillation, a technique where a smaller "student" model is trained on a bigger one to replicate its behavior, to reverse-engineer the model and bring China's capabilities closer to American ones.

The White House hasn't confirmed the report. Trump advisor David Sacks posted on X about the restrictions but focused elsewhere, not China. That silence is worth noting — the story is still a report, not a confirmed move.

What's happening here is a new kind of competition. It's not just who builds the best AI, but who controls access to it. If China can get inside Anthropic's models, it could copy, adapt, and iterate faster than the US expects.

Why this matters for us: Every new AI model that lands in the hands of foreign governments becomes a tool for surveillance, trade, and power — and the communities most affected by US foreign policy are often the ones whose lives get reshaped by these deals.