Judge lets the Claude Max usage-limits suit proceed: core claims survive, others dismissed
A magistrate judge ruled November 6 that Anthropic must face the core Claude Max limits claims: 5x/20x ads may mislead where weekly caps deliver ~3.5x/6x.
A federal magistrate judge has ruled that Anthropic must face the core of a proposed class action over its Claude Max usage limits. On November 6, U.S. Magistrate Judge Alex Tse of the Northern District of California allowed claims of negligent misrepresentation and breach of contract to proceed — the suit by Karl Kahn and Solomon Smith argues that plans advertised as “5x” and “20x” more usage than Pro actually deliver about 3.5x and 6x once weekly caps enter the picture, for at least one Claude model. It is a ruling on a motion to dismiss, not a verdict.
Key points
- The claims that survive: negligent misrepresentation and breach of the subscription agreement. Solomon Smith’s claim under California’s Consumers Legal Remedies Act (CLRA) also proceeds — the judge found Claude Code “could qualify as a service” and his coursework, coding projects and an IBM-sponsored hackathon are not commercial use.
- The numbers in dispute: claude.com markets Max 5x ($100/month) as 5x Pro’s usage and Max 20x ($200/month) as 20x. The plaintiffs calculate the multipliers apply only to each five-hour session; spread over weekly limits, they come to roughly 3.5x and 6x.
- Dismissed: Karl Kahn’s CLRA claim (he may re-plead with detail), two claims under California unfair-competition and false-advertising statutes, and a fair-dealing claim over limits tightened mid-billing-cycle.
Background
The case — captioned Kahn v. Anthropic — was filed in June 2026 in the Northern District of California and first reported by The Wall Street Journal: subscribers said the advertised usage far exceeded what they actually got, and that an ordinary user can hardly measure the allowance they receive. The plaintiffs want refunds for Max subscribers and class treatment; Anthropic declined to comment at the time (via IT之家’s recap of the WSJ report). Anthropic’s consumer lineup runs free, Pro at $17–$20 per month, Max 5x at $100 and Max 20x at $200, with usage governed by two layers — rolling five-hour session windows and weekly caps. That two-layer structure is the technical footing of every number in dispute. The timing is delicate too: Anthropic is reportedly starting its IPO roadshow next week, so a company heading toward public-market fundraising now faces its own subscribers alleging misleading advertising.
What the ruling covers
The November 6 order answers exactly one question: which claims continue. Negligent misrepresentation and breach of contract survive because Anthropic’s terms direct customers to “the order process,” making the sales pages plausibly part of the subscription contract. Smith’s CLRA claim survives on its own logic: Claude Code “could qualify as a service” under the statute, and a computer science student using Claude for coursework, coding projects and an IBM-sponsored hackathon is not a commercial user. Three groups went out. Karl Kahn’s CLRA claim failed because he described his Claude Code use only as “to accomplish computer programming tasks,” without specifics — he can re-plead with detail. The unfair-competition and false-advertising claims failed because the plaintiffs remain subscribers and did not show damages would be an inadequate remedy. And the fair-dealing claim over mid-cycle limit changes failed because the terms permit changes, with no sign of a bait-and-switch. The order’s central line: “The advertising in question is plausibly misleading,” Alex Tse wrote, noting that Anthropic’s per-session disclosure sits “two hyperlinks away from the primary sales pages” on claude.com.
The numbers
How the advertised multipliers compare with the plaintiffs’ math — allegations from an untested complaint, worth holding onto while reading the table.
| Plan | Monthly price | Advertised usage | Plaintiffs’ weekly-cap estimate |
|---|---|---|---|
| Max 5x | $100 | 5x Pro | ~3.5x |
| Max 20x | $200 | 20x Pro | ~6x |
| Pro (baseline) | $17–$20 | — | — |
The gap between those last two columns is the whole case. The 5x and 20x apply only inside each five-hour session; weekly caps do not scale by the same factor. Courthouse News cites a customer example from the complaint: a 20x subscriber burned through half a weekly allowance in three days. Billed monthly, capped weekly — the judge seized on that mismatch: customers who pay by the month could reasonably read the promise as monthly usage.
What others say
Anthropic’s defense runs on a single argument: customers should understand “more usage” as per-session limits, and the company pointed to its help page. Judge Alex Tse was unpersuaded. The plaintiffs had compared the disclosure to shampoo-bottle fine print — useful to read later, but no defense at the point of sale — and the judge noted the disclosure sits “two hyperlinks away from the primary sales pages” on claude.com; subscribers never needed that page to check out. After the ruling, neither Anthropic nor the plaintiffs’ representatives responded to Courthouse News’ requests for comment, and the company likewise declined to comment when the suit was filed in June. The Next Web adds a footnote: Anthropic renamed the Max plans the week before the ruling, so the product line was already in flux.
Our take
For subscribers, the practical lesson arrives before any verdict does: the number that actually binds you is the weekly cap, not the multiplier on the sales page. Heavy users — anyone running Claude Code as a daily workhorse — should reconcile the five-hour and weekly figures against their own burn rate before paying $100 or $200 a month; the page explaining all this is the one the judge said sits two clicks away. For the industry, the order treats the distance between the ad page and the help page as a load-bearing fact: quoting multipliers per session while billing per month now carries legal risk in California. Keep the scope in view, though — this is a magistrate judge at the motion-to-dismiss stage, the plaintiffs still must prove actual harm and causation, and several claims are already gone. The backdrop sharpens the stakes: the roadshow starts next week, enterprise AI spending growth has stalled per Ramp’s data, and Reuters reported on November 5 that potential investors worry an AI backlash could threaten the IPO’s timing. A consumer-trust suit is not the headline Anthropic wanted in that window.
What to watch
The next steps are all checkable. Karl Kahn can re-plead his CLRA claim with specifics; the case then heads toward discovery, where Anthropic’s sales copy, internal pricing and limit documentation could all be pulled. Three points to watch: any motion for class certification, whether Anthropic seeks review of the order, and whether the sales pages quietly drop the multipliers. No court schedule had been published by press time, and Anthropic had not responded to a request for comment.