Anthropic faces legal and customer backlash over misleading Claude subscription claims

A class-action lawsuit and online criticism challenge Anthropic’s claims about the usage capacity of its high-tier Claude plans, raising questions about transparency in AI subscription models.

Anthropic’s most expensive Claude subscriptions are facing scrutiny on two fronts at once: a proposed class action in California and a fresh round of online criticism over what the company’s “Max 5x” and “Max 20x” labels really buy. As of this week, Anthropic’s own pricing and help pages say those plans provide five or 20 times Pro usage “per 5-hour session”, not across a whole week, and also warn that paid plans are subject to weekly limits on top. Court records show the case was filed on 14 June 2026 and amended on 20 July. (anthropic.com)

The claimant, Karl Kahn of Washington, DC, is seeking refunds and class-action status for customers who bought the Max tiers, arguing that the plans delivered materially less usable access than the marketing implied. According to reporting by The Next Web and Benzinga, Kahn says he upgraded to the $200 Max 20x plan in April and found that one five-hour coding session consumed about 15% of his weekly allowance. Anthropic declined to comment to both publications, and the allegations have not been tested in court. (thenextweb.com)

The dispute turns on arithmetic that Anthropic itself published when it explained its new rate limits last year. TechCrunch reported in July 2025 that Pro users could expect 40 to 80 hours a week of Sonnet 4, while Max 5x users could expect 140 to 280 Sonnet 4 hours and Max 20x users 240 to 480. On those figures, the $200 plan works out to roughly 1.7 times the Sonnet capacity of the $100 plan, not twice as much, while Max 5x spans about 3.5 times Pro at the high end and Max 20x about six times Pro at the low end. That gap between the label and the weekly ranges is central to both the lawsuit and the wider customer backlash. (techcrunch.com)

The arguments now being aired in court and on Reddit did not emerge from nowhere. In mid-July 2025, Claude Code users reported that access had become more restrictive without warning, with complaints concentrated among heavy users, many of them on the $200 Max plan. One user told TechCrunch that the change “stopped the ability to make progress”. At the time, Anthropic acknowledged that some Claude Code users were seeing slower responses and said it was working on the problem, while SiliconANGLE linked the complaints to a run of infrastructure disruptions affecting Claude services. (techcrunch.com)

Anthropic then formalised the restrictions. On 28 July 2025, it told subscribers that new weekly limits would start on 28 August for Pro, Max 5x and Max 20x, while the existing five-hour rolling limits would stay in place. TechCrunch reported that the system would include one overall weekly cap and another tied specifically to Opus usage, and that Max users could buy extra capacity at standard API rates after hitting the included allowance. The company said the move reflected “unprecedented demand since launch”, that some customers had been running Claude Code continuously or sharing accounts, and that fewer than 5% of subscribers would be affected. TechCrunch also noted at least seven Claude Code outages in the previous month. (techcrunch.com)

What has changed since then is not the existence of metering, but the clarity of the wording. Anthropic’s current pricing page says Max buyers can choose “5x or 20x more usage than Pro”, with an asterisk noting that usage limits apply. Its FAQ and support articles go further, stating that Max offers 5x or 20x more usage than Pro per five-hour session, that weekly limits are layered on top, that all activity across the web app, desktop, mobile and Claude Code draws from the same pool, and that paid users can turn on usage credits to keep working at standard API prices once they hit the included cap. In other words, the company now describes the product in much more qualified terms than the shorthand that sparked the dispute. (anthropic.com)

Anthropic has also tried to relieve the underlying capacity pressure. In a post published on 6 May 2026, the company said it was doubling Claude Code’s five-hour rate limits for Pro, Max, Team and seat-based Enterprise plans, removing peak-hours reductions for Pro and Max, and expanding compute through a SpaceX data-centre agreement that it said would add more than 300 megawatts of capacity and over 220,000 Nvidia GPUs within the month. That may have improved raw access, but it did not settle the more basic consumer question now before the court: whether a usage multiple is misleading if the practical ceiling depends on a second set of weekly caps and opaque token accounting. (anthropic.com)

For now, the case remains at an early stage. The federal docket shows no ruling on the substance of Kahn’s claims and no certified class. It does, however, show the dispute broadening: Kahn filed an amended complaint in July, Anthropic appeared through counsel, one related-case motion was granted in early August, another was denied on 12 August, and an initial case-management conference is scheduled for 18 September 2026, with a joint statement due on 11 September. That means the viral Reddit argument has already outgrown social media. It is becoming a legal test of how AI companies explain variable, compute-constrained subscriptions to paying customers. (dockets.justia.com)