Claude review: Anthropic's plans, limits and refund terms, read closely

What Claude is and how the plans stack up

Claude is Anthropic's assistant, sold to individuals as Free, Pro, Max and Ultra, and to organizations as Team and Enterprise. We did not test it. What follows is a reading of the pricing page, the Consumer Terms of Service, and Anthropic's help and privacy centers as they stood on November 4, 2026.

The ladder is easy to follow. Pro ($20 a month, or $200 a year) gets you Claude Code, Projects and Research. Max ($100) gives five times Pro's usage. Ultra ($200) gives ten times Pro's weekly usage and, for verified accounts, removes the 5-hour cap. Max and Ultra also include monthly API credits for building on Anthropic's developer platform: $100 and $200 respectively.

The renewal rule to put in your calendar

The clause most likely to cost you money is in the Consumer Terms: cancel at least 24 hours before the last day of your term, or the next period is charged. The terms give an example. Subscribe on January 25 and you must cancel by February 23. If you miss that cutoff, the terms say fees are not refunded, though you keep access until the period ends.

The refund policy itself is short. All payments are non-refundable unless the terms say otherwise or the law requires it. Customers in the EEA and UK can get a refund within 14 days, prorated by use. Residents of Brazil, Mexico, South Korea and Taiwan have a 7-day right to cancel. For everyone else, the help center offers an eligibility check through the support messenger but promises nothing. Upgrading from an individual plan to Team is the one case where Anthropic issues a prorated refund automatically.

How the usage limits work

Anthropic does not sell messages. It sells a share of compute. On paid plans that share resets every five hours, with a weekly ceiling on top. Max and Ultra also have a separate weekly cap for Fable, Anthropic's top model. How quickly you hit a limit depends on the model, the length of the conversation and the size of your files. Claude Code shares the same pool as chat, so a long terminal session eats into the allowance you would otherwise use for chat.

The help center also states that Anthropic may limit usage in other ways, including weekly and monthly caps or limits on specific models and features, at its discretion, and it does not guarantee uninterrupted service. If you need more, you can enable usage credits, which bill at standard API rates against a prepaid balance and a monthly spend cap you set. Those credits don't expire in most countries. The monthly API credits on Max and Ultra are a different bucket: they work only on the developer platform and expire at the end of each billing cycle.

What happens to your conversations

Training on consumer accounts depends on a setting called "Help improve Claude" under Settings, Privacy. Read the retention numbers before you decide:

• with training allowed, de-identified chats can stay in training pipelines for up to 5 years;
• deleted chats disappear from history at once and from back-end storage within 30 days;
• chats flagged by safety systems are kept up to 2 years, and their classification scores up to 7 years;
• feedback you submit with the thumbs buttons is kept for 5 years.

Switching training off stops future use but cannot remove data from models already trained. Incognito chats are excluded from training whatever the setting. For work data, Team plans don't train on your content by default.

Ownership of what Claude produces

The Consumer Terms assign to you whatever rights Anthropic holds in the outputs, as long as you follow the terms, and you keep your rights in what you put in. Nothing in the terms restricts commercial use to paid plans or adds a watermark. The terms also say you should not rely on outputs without checking them, which matters more than ownership for most business uses.