While the big Western labs wall off their frontier models ever more tightly, China is moving the other way. Alibaba has released its new Qwen 3.8 Max — a 2.4-trillion-parameter model, nearly as large as Kimi K3 — as open weights. That’s a reversal: back in May, the company chose not to release its predecessor, Qwen 3.7 Max. And that move has kicked off a genuinely interesting policy debate.
The proposal
Analyst Ben Thompson, in “Who’s Afraid of Chinese Models?”, argues the US should pass a law that establishes two things: first, that collecting data for training is explicitly fair use; second, that terms of service which forbid distillation are unenforceable, at least for US companies.
His logic: distillation — querying a model’s API to train your own — is nearly impossible to stop. Rather than fight a losing battle, the US should lean the other way: a copyright policy that indemnifies the labs and guarantees that what they learned fuels further innovation. He also puts his finger on a sore spot — labs outlaw distillation against their models while having trained on unlicensed data themselves.
The China context
Thompson suspects Alibaba’s decision was also shaped by a speech from Xi Jinping, who in mid-July urged people to “seize this rare, historic opportunity to encourage open source, openness, collaboration and sharing.” If China makes open weights a strategic default while the West closes up — who ends up shaping the global ecosystem?
My take
To me the debate matters more than whether you buy Thompson’s exact prescription. Because it exposes a real tension: openness accelerates innovation and adoption — but weakens the protection labs use to justify their billion-dollar bets. And the other side has arguments too: writing fair use into law wholesale devalues the work of the very people whose data made the training possible. Safety concerns about freely available frontier models haven’t gone away either.
What nags at me is the geopolitical tilt. If open models increasingly come out of China while the strongest Western models vanish behind API walls, this stops being a purely legal question. Maybe the real takeaway is simple: openness has long since become a strategic tool — and both sides now understand exactly how to wield it.
Sources: