Headlines said 'Asian startups clone Mythos.' One is Japanese, Khosla-backed, and the 'clone' claim is unproven
Two weeks after Washington banned Anthropic from exporting its Mythos model, a TechCrunch headline framed two Asian releases as Mythos knockoffs. The thread untangled it: one is a Tokyo lab founded by ex-Google researchers, and nobody has shown either model is distilled from Anthropic at all.
After the Trump administration blocked global access to Anthropic's Mythos, TechCrunch grouped a Tokyo startup's model and a Chinese one as 'Mythos-like' clones filling the gap.
A CTO in the thread pulled it apart: 'Asian' lumps a Khosla-backed Japanese ally in with China, and the 'distilled from Anthropic' line is unproven — these are different multi-model architectures. A skeptic added the obvious tell: you don't ship a frontier model out of nowhere with nothing released before it.
What the export ban set off
Two weeks before this story, the Trump administration banned Anthropic from offering global access to its cyber-capable Mythos and Fable 5 models, deeming them too powerful to export. TechCrunch's piece reported the predictable consequence: other labs rushing into the gap. Tokyo's Sakana AI — founded in 2023 by ex-Google researchers Llion Jones, David Ha and Ren Ito — launched Fugu, pitched as standing "shoulder-to-shoulder" with Fable 5 and Mythos Preview and explicitly sold as a hedge against export controls. China's 360 unveiled Tulongfeng for vulnerability detection, which its founder framed as a "national strategic asset."
The headline's framing was the first thing to go
The category in the headline — "Asian AI startups" launching "Mythos-like models" — collapses under the first expert reply, from a CTO who builds AI systems:
"'asian' is bad wording. this is a japanese startup backed by khosla ventures. japan is an ally of west. the title makes it sound like a chinese company did this." — the CTO of an AI startup, IIT Delhi
The "Mythos-like" part fared no better. Someone who read what the two models actually are pointed out they aren't even the same kind of system:
"Both of the mentioned models are model orchestrators using a vastly different multi model paradigm. Saying they in particular are distilled from Anthropic is really [citation needed]." — a commenter on the models' architecture
And the cleanest piece of skepticism was about provenance — frontier models don't appear from nowhere:
"Has either of these companies released models before this? … Deepseek, Z.ai, Alibaba/Qwen have been at this for a lot longer … I find it hard to believe that these new companies would just suddenly release a Mythos-level model without releasing anything prior." — a commenter with \~11k karma
The grievance under the geopolitics
The flamewar the headline invited did show up — enough that a moderator stepped in — but the recurring substantive note was about how the training data got there in the first place:
"they mined the internet first. now they're upset someone brought a shovel." — the same AI-startup CTO
That's the reversal worth keeping. The story sold as "rivals clone America's dangerous model" turns out to be one allied lab and one Chinese lab shipping different architectures, with the distillation claim resting on nothing public — and an export ban that mostly advertised the thing it was meant to contain.
Get breakdowns like this every day
The top Hacker News stories, unpacked — by email, Telegram, or RSS.
A Kindle owner reopened a novel he'd bought. It had quietly become the German translation
He'd paid for Stephen Baxter's 'Ring' years earlier. On a new device it redownloaded as the German edition — Amazon called it a glitch, refused to fix it, refunded him, and left him unable to read the book he owned.