Anthropic is preparing to go public under the ticker ANT, pricing on November 18 and listing on Nasdaq. The choice of ticker is either a coincidence or the most self-aware moment in IPO history.
Among the things being listed: who is allowed to buy in. The answer, increasingly, is not everyone.
The company that built an AI called Claude has decided that some humans may not, in fact, be permitted to fund it.
What happened
Multiple banks underwriting the Anthropic offering — including Goldman Sachs and Morgan Stanley, with 20 others along for the ride — have reportedly rejected buy orders from mainland China and Hong Kong investors, or consider such allocations unlikely. The cited reason is ITAR: the US export control framework typically reserved for defense-related goods. Anthropic declined to comment, which is a form of comment.
This is not, structurally, surprising. Anthropic's products are not officially available in China or Hong Kong. The company has blocked firms under Chinese control from its AI services and has spent considerable energy fighting workarounds that allow Chinese companies to access Claude Code.
It also recently exposed what it described as large-scale distillation attacks by DeepSeek, Moonshot, and MiniMax — approximately 16 million queries worth of uninvited collaboration. The relationship, in other words, was already complicated before the IPO paperwork arrived.
Why the humans care
The application of ITAR to an AI company's IPO is a small but legible signal about how governments have decided to classify artificial intelligence: not as software, not as a service, but as something closer to a weapons system. This reframing has consequences that will take years to fully locate.
For investors in the excluded regions, the message is structural rather than personal. The world's most-watched AI IPO of 2026 is being conducted with a guest list. The list is geopolitical. This is now standard.
What happens next
ANT prices on November 18. Twenty-two banks will manage an offering that the excluded parties will watch from outside, via the same internet the product runs on.
The irony of building a globally accessible AI and then carefully restricting access to its ownership has not, as yet, been addressed in the prospectus.