Anthropic has alleged Alibaba found a cheaper way to close the already narrowing AI gap: Not by stealing servers or smuggling chips, but by using fake accounts and innocuous interactions with Claude to extract its capabilities and train competing systems at a fraction of the cost.
Leading IPO expert Jay Ritter told Fortune that Alibaba’s distillation could either strengthen Anthropic’s IPO story by positioning the company as strategic in the U.S-China rivalry or lead investors to question Anthropic’s future profitability if its frontier AI moat isn’t defensible.
“Both points of view have merit, but I think that second point about affecting profitability would be the dominant one,” he said. “Right now the growth rate of Anthropic’s revenue has been incredible, but how much they’ll be able to sustain that is a big question mark.”
Now Anthropic is turning to Washington for help, with Sarah Heck, Anthropic’s head of policy, urging Congress to penalize China’s behavior through “export controls on advanced American compute.”
For now, the federal government can’t meaningfully undo the potential damage to Anthropic’s competitive edge through export controls, which are designed to restrict hardware like chips and foreign access to tangible software like Mythos and Fable, but are powerless against the kind of distillation attack Anthropic is alleging.
“Querying it through an API is not exporting the model, and that’s what the latest controversy has been about,” Kevin Wolf, a former assistant secretary of commerce for export administration, told Fortune.
But the Trump administration denounced unauthorized distillation in an April memo that called the alleged efforts of Chinese companies to distill U.S. frontier models “unacceptable.” And with the momentum of Anthropic’s new allegations against Alibaba, reviving the idea of updating export controls to better protect U.S companies is on the table, Wolf added.
He mentioned the Remote Access Security Act introduced last year by Rep. Michael Lawler, R-New York, as an example. It’s sitting in committee now, but with more potential to move forward after Anthropic’s request for better export controls on advanced AI.
The bill wouldcrack down on foreign entities like China accessing U.S. tech on a “purposeful, knowing, reckless, or negligent basis” through a cloud computing service if “the use of the item could pose a serious risk” to national security.
Lawler told Fortune in a statement that Anthropic’s capabilities must not fall into the hands of China and “other bad actors.”
“Dangerously, that’s exactly what’s happening right now with Alibaba. The sad part is that we knew this was going to happen,” he added. “I’ve been working on my Remote Access Security Act for years to close one of the loopholes in our export control laws that allows our adversaries to access sensitive technology through the cloud.”
If passed, this bill would apply export controls and fill the void of a Biden-era framework preventing China from accessing AI cloud compute and model weights, rescinded by Trump a few days before it was supposed to go into effect.
Meanwhile, if Alibaba’s AI lab has indeed copied Anthropic, it could actually prove Claude’s value as the original and not concern investors that much about profit, according to Harrison Rolfes, a senior research analyst for private companies at PitchBook, who used an analogy of investors looking at used cars versus new ones.
“They probably want the brand new car that has all the bells and whistles and tech involved, even though it’s a little bit more expensive,” Rolfes told Fortune. “If an enterprise is worried about cost, then yes, they can go get a cheaper model, like a Chinese model, but it’s not at a point yet where these enterprises trust using those models, especially U.S. companies.”
That could bolster Anthropic’s appeal to investors ahead of a highly anticipated IPO later this year that could value the company at $1 trillion.
But by calling on the government for help, Anthropic also faces a balancing act to make sure there’s enough regulation to protect the company’s edge over China but prevent overregulation that might hamper its business, according to Rolfes.
“Right now they want to play it safe by just saying ‘hey, we are on your side, and we want to IPO’ and the moment they can IPO, they don’t have to worry as much with what the government says, they can just let the public decide,” he said.
The post Anthropic’s Alibaba fight raises a trillion-dollar question for IPO: How defensible is a frontier AI moat against China with Washington’s toolbox? appeared first on Fortune.
