BotBeat
...
← Back

> ▌

AnthropicAnthropic
FUNDING & BUSINESSAnthropic2026-07-11

Anthropic Faces Billing System Crisis: $16.6M Phantom Invoice Charges Korean User

Key Takeaways

  • ▸A Korean Claude user received $16.6M in phantom invoices from Anthropic within 24 hours, with zero recorded API usage and no registered payment method
  • ▸Invoices appeared legitimate and routed through official Anthropic/Stripe infrastructure, but the billing system attempted to charge a card the user claims was never registered
  • ▸No actual money was collected—both charge attempts were declined due to transaction limits—but the incident raises serious questions about Anthropic's billing security and controls
Source:
Hacker Newshttps://www.internationalcyberdigest.com/anthropic-tried-to-phantom-charge-16-6m/↗

Summary

A Korean Claude user reported receiving phantom invoices from Anthropic totaling $16,627,739.70 in early July, allegedly generated by the company's Stripe billing system within 24 hours. The user claims to have been on the free tier with zero API usage and no registered payment card, yet received two invoices via legitimate-appearing Stripe emails from Anthropic's official domain—first for $1.67M, then $16.6M.

While both invoices ultimately failed to collect payment after being declined by the user's card issuer, the incident reveals potential vulnerabilities in Anthropic's payment processing infrastructure. Bank statements show two charge attempts made to a card the user insists was never registered on his account, raising questions about how the billing system obtained payment information and initiated charges without proper authorization.

Anthropic has declined to provide details to journalists at ZDNet, stating only that individual account matters are confidential. The company has not yet clarified whether the phantom invoices resulted from a billing error, security breach, or system malfunction. As of publication, the underlying cause remains unexplained, and the user has indicated he will share updates as the situation develops.

  • Anthropic has refused to comment on specifics, leaving open questions about root cause and whether other users may have experienced similar issues
MLOps & InfrastructureCybersecurityPrivacy & Data

More from Anthropic

AnthropicAnthropic
RESEARCH

Anthropic Unveils Hidden 'J-Space' Inside Claude Using New Mechanistic Interpretability Technique

2026-07-11
AnthropicAnthropic
RESEARCH

Argument Collapse: Research Reveals LLMs Generate Dramatically Less Diverse Arguments Than Humans

2026-07-11
AnthropicAnthropic
RESEARCH

Local-First Agent Governance: A Framework for Safe, Contained AI Agents

2026-07-11

Comments

Suggested

Nebula SecurityNebula Security
RESEARCH

AI-Driven Tool Discovers 15-Year-Old Linux Root Vulnerability

2026-07-11
AnthropicAnthropic
RESEARCH

Local-First Agent Governance: A Framework for Safe, Contained AI Agents

2026-07-11
Google / AlphabetGoogle / Alphabet
PRODUCT LAUNCH

Google Launches LiteRT.js: Native-Speed AI Inference Comes to the Web

2026-07-11
← Back to news
© 2026 BotBeat
AboutPrivacy PolicyTerms of ServiceContact Us