Hidden Limits Behind Premium Tiers
AI labs face mounting pressure to subsidize punishing inference costs, and the cracks in their packaging are showing up in court. A group of power users has filed an expanded class action lawsuit alleging that Anthropic engaged in deceptive advertising regarding the throughput limits of its premium Claude Max tier. The legal challenge targets the gap between marketed capacity and operational reality for demanding business workflows.
Anthropic prices Claude Max in two tiers: $100 per month for a advertised 5x multiplier over the standard Pro limits, and $200 per month for a 20x multiplier. According to the complaint, Anthropic applies these multipliers exclusively across rigid five-hour rolling windows that are quietly capped by an overriding weekly ceiling. Power users and engineering teams discover that this compounding architecture yields substantially less usable compute than headline marketing suggests.
The Legal Dispute Over Fine Print
Former Federal Trade Commission attorneys Monica Vaca and Kati Daffan—both veterans of FTC enforcement under Lina Khan—are spearheading the litigation against Anthropic.
As attorney Monica Vaca pointed out, subscribers typically upgrade during mission-critical sprints, only to encounter opaque throttling mechanics and artificial bottlenecks buried behind opaque interface prompts.
Hyperlinks and Shifting Terms
Anthropic rolled out the Max plan in April 2025 and introduced the contentious weekly throttling caps shortly after. As detailed in the legal filing, deciphering the actual throughput terms requires parsing layered hyperlinks scattered across disparate support pages just to locate the operational definition of a single usage session.
Selling capacity multipliers while masking hardware strain behind dynamic throttling and buried quotas is becoming an untenable product strategy. For technical leaders and enterprise teams, consumer-grade 'unlimited' or tiered seat plans remain a reliability risk; production pipelines require hard API contracts with guaranteed tokens-per-minute (TPM) and explicit throughput SLAs rather than moving marketing targets.