Two Ex-FTC Lawyers Expand the Claude Max Class Action: Anthropic's '5x/20x' Marketing Meets False-Advertising Law

Anthropic likes to say power users are the core of its business — it has prioritized them even when doing so meant cutting off entire categories of popular applications, most publicly the OpenClaw agent platform. On September 8, some of those same power users struck back in federal court with an expanded class action that accuses the company of deceptive advertising on its most expensive consumer subscription tiers, and the attorneys leading the case come with a very specific pedigree: Monica Vaca and Kati Daffan spent a combined 38 years at the Federal Trade Commission under chair Lina Khan.

The lawsuit, first filed in July and withdrawn to be refiled today as an expanded class action, targets the Claude Max plans — $100 per month for “5x” the usage of the $20 Pro plan, or $200 per month for “20x.” The core allegation: those multipliers, splashed across Anthropic’s marketing graphics, are quietly hollowed out by a structure of five-hour rolling session windows layered under a weekly cap. The complaint argues the real increase in usable capacity adds up to far less than the headline numbers suggest.

The anatomy of a “5x”

To see the problem the plaintiffs describe, you have to reconstruct what a subscriber actually buys. The Pro plan at $20 per month comes with a defined pool of usage. Max 5x promises five times that pool; Max 20x promises twenty times. But according to the complaint, the multipliers apply only within five-hour “sessions” that are themselves subject to a weekly ceiling. A user who burns through their weekly allotment by Wednesday gets nothing more until the window resets, no matter how large the advertised multiplier was.

Monica Vaca, in an interview with The Verge, described the disclosure structure as requiring users to click through two separate hyperlinks before ever encountering the word “session” — and then to navigate to a different page entirely, the Pro plan documentation, to learn what “session” actually means. “This is hard for consumers — they don’t know what’s in that black box,” she said. “There’s no way for them to audit it. They’re basically taking a leap, and they’re believing in an honest marketplace.”

The frustration is not hypothetical. A Reddit user quoted in coverage of the case put it in mechanical terms: “The weekly allowance is what the pricing page makes you think you’re buying. The rolling 5-hour window is what actually controls whether you can work … It’s like giving someone a bigger gas tank while keeping the fuel pump limited to one gallon every five hours.”

Anthropic’s defense: read the back label

Anthropic has already responded once. In a motion to dismiss the original July complaint, the company did not deny that the session and weekly limits exist — it argued that the information was technically available to anyone who went looking. “Accessing this clarifying information … required nothing more than clicking hyperlinks available in the purchase process — the digital equivalent of flipping a product over to read the back label,” Anthropic wrote in the filing.

That framing does two things at once. It concedes the substance of what users experienced — the caps are real, the multipliers are conditional — while contesting the legal standard for when buried terms become deception. U.S. false-advertising precedent draws a line between terms a reasonable consumer would notice and terms engineered to be missed. The plaintiffs’ argument is that a “20x” claim whose real-world effect depends on definitions scattered across two pages fails that test; Anthropic’s is that the consumer’s browser had a back button.

The company declined to comment on the expanded complaint. That in itself is a shift: earlier this year it publicly defended its pricing changes, and in the release notes for Claude Fable 5.1 it acknowledged customer price sensitivity, writing that it was “addressing the feedback we’ve received from customers on price” through the new model’s tiered pricing.

Why now: the industry’s money squeeze

The timing of the expansion is not accidental. Over the past eight months, as pressure to turn a profit has intensified across the frontier labs, customers have complained repeatedly that the top AI companies are passing their steep operating costs downstream — through throttling, quieter caps, and repricing of once-generous tiers.

The Max plan launched in April 2025 without the contested weekly limits. According to the complaint, those limits arrived a few months later, in August, as Anthropic pushed to compete with OpenAI on margins rather than just benchmarks. In other words, the product subscribers signed up for and the product they ended up with diverged mid-subscription — a pattern that consumer-protection law treats with particular suspicion, because the advertised terms were true when the subscription started and quietly stopped being true while the price stayed fixed.

There is also a competitive backdrop the lawsuit implicitly gestures at. OpenAI’s ChatGPT Pro and its Codex tooling have gone through their own cycles of generous limits followed by retrenchment, and both companies face the same structural math: agentic workloads consume tokens at rates flat-rate subscriptions were never designed to absorb. Claude Code users on weekly caps, API customers watching effective throughput, and Max subscribers hitting invisible ceilings are all feeling the same underlying economics.

The ex-FTC angle

The most consequential fact about this filing may be who signed it. Vaca and Daffan are not generic class-action attorneys — they are former federal consumer-protection enforcers who built careers on exactly this category of case. “There is a long line of precedent on false advertising,” Vaca told The Verge. “It’s commercial speech. You can’t lie when you’re marketing a product.”

That experience shows in the theory of the case. Rather than framing the dispute as a contract quarrel over ambiguous terms, the complaint frames it as a false-advertising claim about what a reasonable consumer would take from the pricing page. That framing matters because it routes around the arbitration clauses and terms-of-service defenses that AI companies have used to deflect consumer suits. False-advertising and consumer-protection statutes are designed to reach exactly this pattern: a prominent number, a qualified reality, and a gap the customer could not audit.

Vaca also described a second motivation that goes beyond the mechanics of session windows. She and Daffan started hearing from workers who felt they had to maintain expensive AI subscriptions to stay relevant in the job market — and who then found the service delivered less than advertised. “People see that they are going to get this dramatically expanded usage,” she said. “What we hear from people, though, is that when they sign up, they’re surprised that they’re not getting the usage that they thought they were getting.”

What to watch

Three things will determine whether this case matters beyond its docket number.

First, the motion to dismiss. Anthropic’s “back label” argument worked well enough to get the original complaint withdrawn and rewritten; the expanded version is presumably built to survive it. If the case clears dismissal, discovery into Anthropic’s internal conversion and cancellation data — how many Max subscribers hit the weekly cap, how quickly, and what the company knew about the gap between marketing and reality — could be revealing.

Second, whether the theory generalizes. Nearly every major AI subscription now sells a multiplier or an “unlimited-ish” promise qualified by rate limits, deprioritization windows, or fair-use policies. A ruling that conditional multipliers require clearer disclosure would bind far more than one pricing page.

Third, the precedent it would set for the IPO-bound labs. Anthropic is expected to go public at a valuation north of $900 billion in the coming months, and consumer-litigation risk over subscription economics is precisely the kind of line item that shows up in the risk factors section of an S-1. A certified class of Claude Max subscribers is a contingent liability with a number attached.

For now, the case is at its earliest stage, and Anthropic has lost nothing but the argument in the court of public opinion. But the substance of the dispute — a company selling “20x” while operating “20x, conditions apply” — is one the entire subscription-AI industry should recognize, because every lab running metered compute under a flat-rate price is one quiet policy change away from the same docket.