Anthropic has been grifting its highest-paying users, a new class action lawsuit alleges.
Filed Monday in the Northern District of California and first reported by the Wall Street Journal, the lawsuit claims that Anthropic has been misleading consumers with its marketing for Claude Max 5x and Max 20x subscription plans. Respectively, the plans cost $100 and $200 per month and are supposed to come with five and twenty times the token allowances as the less expensive Claude Pro subscription. But the plaintiff behind the lawsuit, a Washington D.C.-based user named Karl Kahn, argues the actual limits seem to be much lower, and Anthropic’s pricing model makes it difficult to determine exactly how and where users’ tokens are being spent.
The lawsuit claims that Anthropic sent emails to Max 5x and Max 20x subscribers in July of last year outlining the token allowances they could expect every week, and that those were “far below the advertised amount of usage,” according to a quote from the filing included in the Journal’s report. Anthropic declined to comment on the new lawsuit.
Costs and complaints
Khan’s allegations are the latest in a flurry of complaints against the swelling costs of using frontier AI models. Companies like Amazon and Uber have recently taken steps to limit employees’ use of AI, citing the technology’s price tag, and OpenAI CEO Sam Altman has called the financial burden of AI “a huge issue” facing enterprise customers. The new lawsuit also underscores the complexity that often comes with trying to translate vagaries like tokens and usage windows into concrete monetary values.
In response to the rising tide of complaints, developers like Anthropic and OpenAI have faced pressure to cut down costs for users, even as the price of training new AI models also continues to climb. Open-source models like DeepSeek have meanwhile been growing in popularity as businesses and individuals get priced out of using proprietary models.
Models have also been getting more token-hungry as they become more “agentic,” or autonomous. Asking Claude to rewrite a work email in a more professional tone doesn’t run through much of a user’s token allowance, but it’s a much different story for complex, long-running tasks, like designing a website landing page, which could require several series of steps, the deployment of sub-agents, and therefore much more computing power. Shortly after the debut of Anthropic’s Fable 5 early last week—a model which Anthropic specifically touted for its agentic capabilities—some users started complaining that they were burning through their usage windows far faster than with earlier models.
An underdog in the AI race, not so long ago, Anthropic recently surpassed OpenAI as the highest-valued startup in the world. Its fame has also swelled following its very public dispute with the U.S. Department of War over the use of its AI in autonomous weapons systems and other military technology. On Friday, the company was ordered by the federal government to cut off access to Fable 5 (along with another powerful model called Mythos 5) for all foreign nationals in the U.S. and abroad, due to alleged cybersecurity concerns.
The affordability angle
Meanwhile, Anthropic’s competitors have rushed to take advantage of the complaints being leveled against the cost of its Claude chatbot.
During its most recent Build developer’s conference earlier this month, for example, Microsoft unveiled a new AI model called MAI-Thinking-1 (along with six others), which company AI lead Mustafa Suleyman said performs comparably to Anthropic’s Claude Opus 4.6 for a lower cost. Suleyman also told Bloomberg that “many people are urgently looking for alternatives” to Anthropic’s expensive models.
Similarly, OpenAI—which, along with Anthropic, is expected to move forward with its initial public offering later this year—appears to be leaning into affordability as a sales tactic as it reportedly considers significant cuts to its current token prices.
Read the full article here
