Anthropic sued over deceptive Claude Max pricing

If you just paid $200 for Claude Max expecting twenty times the firepower of your base plan, only to hit a brick wall three hours into a coding sprint, you are not alone.

Anthropic was hit with an expanded class action lawsuit today over allegedly deceptive marketing on its Claude Max subscription tiers. The suit was filed by Monica Vaca and Kati Daffan, two veteran attorneys who spent a combined 38 years at the Federal Trade Commission, serving under Lina Khan.

It is one of the first major legal challenges targeting a growing grievance in the generative AI space: tech labs squeezing paying subscribers as operating costs spiral.

A bigger gas tank with a broken pump

Anthropic pitches its Max plan as the ultimate upgrade for power users over the standard $20-per-month Pro tier. The pricing is steep: $100 per month for five times the usage limits of Pro, or $200 per month for twenty times the limits.

According to the complaint, those splashy multipliers only tell a fraction of the story. The suit alleges that the 5x and 20x multipliers apply strictly within rolling five-hour session windows, which are also shackled to separate weekly caps. That combination results in dramatically less real-world usage than a straightforward 20x boost would suggest.

Finding that out requires serious digital excavation. Users have to click through two separate hyperlinks just to spot the word session, and then navigate over to the Pro plan documentation to figure out what a session actually means.

One Reddit user quoted in the coverage summed up the frustration neatly: the weekly allowance makes you think you bought an open runway, but the five-hour window dictates whether you can actually get work done. It is basically like buying a car with a massive fuel tank, only to find the gas pump restricts you to one gallon every five hours.

Anthropic first rolled out the Max tier in April 2025, but it did not introduce the controversial weekly caps until August, right as competition with OpenAI heated up and compute expenses mounted.

The fine print defense

In a motion to dismiss an earlier version of the case filed in July, Anthropic defended its setup. The company argued that the session limits were never concealed, insisting the information was readily available to anyone willing to look.

Clicking through hyperlinks during checkout, Anthropic claimed, is simply the digital equivalent of flipping a box over at the grocery store to read the nutrition label on the back.

Former FTC attorney Vaca is not buying that logic. She pointed out that generative AI models are an opaque black box to everyday consumers. Users cannot independently audit their compute consumption or forecast quota burns, meaning they have no choice but to take marketing claims at face value.

Many users upgrade on the fly in the middle of urgent projects precisely because they assume paying $200 will clear their bottlenecks, only to get cut off anyway. Vaca also noted that many subscribers feel intense career pressure to pay top-dollar for these tools just to stay competitive at work, making hidden throttling feel even more predatory.

Why it matters

This lawsuit cuts straight to the core tension of the current AI boom: the economics of cutting-edge frontier models are punishingly expensive, and vendors are quietly offloading that strain onto their most dedicated users.

For power users, developers, and knowledge workers who rely on Claude daily, this case is validation of an open secret. If you are debating whether to upgrade to Claude Max right now, be careful. Unless your workload is spread neatly across evenly spaced five-hour intervals, that $200 tier will likely leave you just as throttled during crunch hours as the standard plan.

Across the industry, labs are scrambling to stem compute burn. Anthropic even severed access for popular third-party applications like OpenClaw to ringfence capacity for first-party subscribers, and subtly acknowledged customer pushback on pricing when introducing its Fable 5.1 model.

We think Anthropic's defense—treating multi-variable AI rate limits like a cereal box label—is remarkably out of touch. Measuring neural network consumption is not like counting calories, and burying restrictive rolling throttles behind nested links looks deliberately confusing.

If this lawsuit succeeds, it could set a major regulatory precedent. The entire AI sector might finally be forced to abandon misleading multiplier marketing and explicitly state clear, upfront token or prompt allowances before asking for hundreds of dollars a month.