AI’s Cost Problem Becomes Front and Center, At Least in Asia

Anthropic can’t catch a break! Leaders of the AI firm are meeting with Trump administration officials in D.C. this week to try to resolve a dispute that blew up on Friday night. The administration said that it would need to cut off access to its most advanced Claude Fable 5 and Claude Mythos 5 models to foreign nationals, which led to the company taking down the models altogether. 

Then, on Saturday, my colleagues and I reported that the latest dispute with the government stemmed, at least in part, from concerns Amazon CEO Andy Jassy had shared with senior Trump administration officials in recent days, regarding security risks in Anthropic’s most advanced models. It’s not clear why Jassy did this, as Amazon is one of Anthropic’s largest investors and suppliers of chips and compute to run and train its models. There could be some 5D chess going on, but it could also be as simple as Jassy not realizing that the situation would escalate to where it did.

It’s not clear what the solution might be, but perhaps Anthropic could try and sweeten the deal by strengthening its model guardrails or offering a stake of the company to the government, something that’s been floated with other AI labs. All this comes as tensions between Anthropic and customers are escalating, as we reported on Thursday.—Stephanie Palazzolo

Onto today’s column…

While AI hasn’t faced a public-relations problem in Asia like in the U.S., it is confronting another challenge: rising costs. And that threatens the popularity of AI services.