Chinese AI Researchers Love Claude — Then Get Banned
Inside China, top AI talent relies on US tools — and resents being locked out.
An essay titled "The Anatomy of a Chinese AI Researcher" has been drawing attention on LessWrong, a forum where AI safety researchers and enthusiasts argue about where the technology is heading. It isn't a study, a leak or a news report — it's a sketch of a character: a scientist at a Chinese "frontier lab" (a company building the most powerful AI systems) who is racing to catch up with Western firms and knows how hard that is. Readers on the forum clearly recognised the type.
The essay's central contradiction is everyday and practical. This researcher depends on American AI tools — Claude from Anthropic, plus OpenAI's models — for his daily technical and intellectual work. But he has now had three paying accounts banned, because China is not a supported region for those services. He vents on Chinese developer forums, where users accuse Anthropic's leadership of hostility toward China and of "safety brainwashing." For a working scientist, having your main tool disappear mid-project isn't an abstraction; it's lost time and lost work.
His deeper complaint is about how he's described. Western AI safety leaders, he argues, discuss people like him only as a source of risk, never as someone who could be trusted to help manage that risk. The essay says effective altruism groups — a movement focused on long-term threats to humanity — never reach out to researchers like him in a culturally appropriate way. So he stays in "capabilities" work, meaning making AI more powerful, because that's what pays and what his employer rewards. He quietly admits he'd rather work on a world where humans and AI flourish together.
Commenters on the post offered blunt advice: stop trying to win hearts and minds, and simply pay above-market salaries to recruit strong Chinese AI researchers into safety work. The takeaway for everyone else is that AI safety is a global problem. If the people building the most advanced systems feel shut out of the conversation about how to build them safely, that conversation is missing half its participants.
- Chinese AI researchers use American AI tools daily, but paid accounts get banned because China isn't a supported region.
- The essay says Western safety groups treat these researchers only as a risk, never as partners — so they keep building instead of safeguarding.
- Commenters' suggested fix is blunt: pay top Chinese AI talent above-market salaries to work on safety.
Why It Matters
AI safety is global — excluding the people building the tech makes everyone less safe.