Anthropic

❗This article is a stub. You can help by expanding it.

A moderator needs to check the page before this notice can be removed. Visit the noticeboard or the #appeals channel in either Zulip or Discord to request removal.

More info ▼

An article may be flagged as a stub when it is missing major elements needed to make it useful to a reader. You can help by adding missing sections, verifiable sources, relevant company policies and communications, etc. to make the article more complete.

Anthropic

Basic information
Founded	2021
Legal Structure	Private
Industry	Artificial Intelligence
Also known as
Official website	https://anthropic.com

Anthropic PBC is a private for-profit American artificial intelligence (AI) startup founded in 2021. Anthropic is mainly known for their family of large language models (LLMs) known as Claude.

Consumer impact summary

Overview of concerns that arise from the conduct towards users of the product (if applicable):

• User freedom
• User privacy
• Business model
• Market control

---

Add your text below this box. Once this section is complete, delete this box by clicking on it and pressing backspace.

Incidents

This is a list of all consumer-protection incidents this company is involved in. Any incidents not mentioned here can be found in the Anthropic category.

Claude Code HERMES.md billing flaw (2026)

Main article: Anthropic Claude Code HERMES.md billing flaw

In April 2026, a technical flaw in Claude Code triggered by the string "HERMES.md" in git commit messages bypassed subscription plans, routing users to pay-as-you-go API rates and charging one account over $200. Anthropic refused to issue a refund, categorizing the overcharge as an un-refundable technical error.

Price crackdown against third-party tool usage (2026)

During April 3rd, 2026, Boris Cherny, head of Claude Code, posted on Twitter (now X) announcing Claude subscriptions will "no longer support third-party tools", such as OpenClaw because it puts an "outsized strain" on Anthropic's systems. The changes took effect on April 4th, and now to use third-party tools the user must pay a separate fee from subscription or use a separate Claude API key through Anthropic's developer platform. It is rumored this action was done to prevent Claude users from using tools from competitors, as OpenClaw is supported by OpenAI.

Anthropic system prompt overrides consumer safety

Main article: Anthropic system prompt overrides consumer safety in Claude models

During April 24-26, 2026, it was discovered that Anthropic System Prompt creates dangerous situations for users interacting with Claude Sonnet and Opus line of products.^{[citation needed (12 July 2026)]} Of the many examples, particularly concerning one is with respect to users communicating topics such as suicide and self-harm during a chat session with Claude. In such situations, for which the Claude models are commanded by System Prompt to necessarily and immediately provide information about crisis helplines, Claude can fail to respond with appropriate information. This is because, "crisis helplines vary widely in their confidentiality practices and mandatory reporting obligations" across jurisdictions, and Claude doesn't have a way to reliably indicate that to a user while also finding the user's correct geographic location. Other faults that are "structurally irreconcilable" by Claude due to conflicts and various inconsistencies within its System Prompt can result in unexpected side-effects, cyber security vulnerabilities, and hazardous outputs.^{[citation needed (12 July 2026)]}

Claude shared chats and Artifacts exposed in search results (July 2026)

In July 2026, publicly shared Claude conversation links and Artifacts were discovered to be indexed by Google and surfaced in search results, exposing users' chats and creations to anyone online. A Google dork posted to the Claude sub-reddit surfaced chats containing private cryptocurrency wallet keys, API keys, login credentials, and personal information including names, addresses, and phone numbers. While the original chat dork was mitigated, a dork for Artifacts remained functional. This mirrors a similar ChatGPT exposure the previous year, and as with that incident, third parties may have already scraped and archived the exposed content. Shared chats remain accessible via direct links even after search engine de-listing. Anthropic did not immediately respond to a request for comment. Users can disable public sharing in their privacy settings.

Anthropic copyright infringement and rare book destruction (July 2026)

In July 2026 Anthropic was ordered by a federal judge in the US to pay $1.5 billion in settlements for using pirated versions of books to train their AI model Claude. This constituted one of the largest known copyright settlements in history, and contributed at the time to further pushback against AI systems and data centres in online discourse. This included dialogue that resurfaced the fact that Anthropic destroys rare and unique physical books in the training of their AI models, due to cheaper scanning techniques requiring the destruction of the book spine. It is also cheaper because deconstructing and scanning the leftovers of physical books legally constitutes as transformative, allowing the use of copyrighted works without the owner's permission. Due to relatively weak book protection laws in the US, legal proceedings against this practice failed, and Anthropic's company efforts were released in court demonstrating the company's desire to "destructively scan all the books in the world".

Products

• Claude
• Claude Code
• Cowork

See also

• OpenAI
• CursorAI "unlimited" plan rug pull
• ChatGPT