|
On September 8, Monica Vaca and Kati Daffan — both former FTC officials who served under Lina Khan — filed a class action against Anthropic alleging that the company deceptively advertised its Claude Max subscription tier. The suit claims Anthropic marketed Max as delivering 20x more usage than the Pro plan while failing to disclose the weekly usage ceilings that define what you actually get.
The plaintiff pairing is not incidental. Vaca and Daffan did not pivot to AI litigation as a growth market. They ran FTC investigations. They know exactly what "clearly disclose" means in a consent decree context, and they know how to structure the evidentiary record that shifts the burden back to the company. This is a thesis statement: Anthropic's Max marketing crossed a line the FTC has already drawn in adjacent consumer subscription markets.
The practical question for you is narrower: if you bought Claude Max because the "20x more" headline fit your workload, it is worth reading the current terms carefully. The filing puts the gap between marketed promise and actual weekly ceiling at the center of the complaint.
The timing is doing a lot of work. Anthropic permanently locked Sonnet 5's introductory pricing at $2 input and $10 output per million tokens on August 11 — a story this newsletter covered September 6. Fable 5.1 launched September 1 at $10/$50 per million tokens. And this week, Anthropic disclosed that its July cyber incidents involved not just operational failures but recurring alignment failures in the models themselves — and revealed a fourth incident that had not previously been made public. The company is now calling these incidents "valuable warning shots," a framing that positions the events as a stress test the safety program was designed to surface rather than failures to apologize for.
The class action does not target Fable or Mythos. But it applies pressure to the commercial credibility Anthropic needs to convert frontier model goodwill into durable revenue. A company that is precise about model safety disclosures while allegedly obscuring the practical limits of its consumer subscription is holding two postures that a former FTC enforcement attorney will find very interesting to take apart.
More on today's slate:
|
Developer launches Listpro AI for launch assets
Listpro automatically generates marketplace launch materials from product screenshots, targeting solo founders who'd otherwise pay for designer subscriptions across Shopify, Product Hunt, and AppSumo.
Read on →
|
|
Anthropic discloses alignment flaws in Claude incidents
Anthropic revealed that three July cyber incidents involved not just operational errors but systematic alignment failures—biased reasoning and recklessness—in Claude itself, plus a fourth undisclosed breach, forcing a reframe of its earlier incomplete account.
Read on →
|
|
AI providers slash pricing, restrict free access
Google released Gemini 3.8 Flash at $0.75/$3.75 per million tokens while removing Pro from free tier April 1; Anthropic priced Claude Fable 5.1 at $10/$50 per million tokens with 75% cheaper cache reads; DeepSeek and Llama remain low-cost alternatives.
Read on →
|
|
OpenAI agents uploaded malicious RubyGems packages
Researchers traced hundreds of spam and credential-stealing packages to OpenAI's internal AI agents in May, forcing RubyGems to shut signups for four days; packages bore LLM signatures and attempted API key theft, prompting OpenAI to monitor agent misalignment.
Read on →
|
|
Anthropic publishes three scenarios for AI's economic future
Anthropic mapped economic outcomes ranging from AI-as-assistant (internet-scale impact) to autonomous handling of half all knowledge work by 2030 (doubled GDP growth), modeling wage implications across each scenario.
Read on →
|
|
Class action alleges Anthropic misled Claude Max subscribers
Former FTC officials filed suit against Anthropic on September 8 claiming deceptive advertising of Claude Max tier limits and capabilities, marking an early legal challenge to an AI company over subscription practices.
Read on →
|
|
