💼 $1.7 Million Billing Errors at Anthropic and OpenAI
A Vaudit audit has revealed overcharges to Anthropic and OpenAI customers. The cause was infinite retry loops during task execution in Anthropic's Claude Code, for which tokens were continuously billed. Affected companies include Panasonic, HP, and Honda.
🌍 The incident highlights the vulnerability of the token-based payment model when working with autonomous agents. There is a need for "infinite loop" control mechanisms and transparent billing for enterprise clients.
👤 Users of AI agents should set limits on the number of attempts and monitor expenses in real-time to avoid unnoticed budget depletion.
Source 1: https://www.businesstoday.in/technology/artificial-intelligence/story/anthropic-openai-customers-overbilled-by-1-7-million-due-to-billing-errors-539626-2026-06-29 Source 2: https://x.com/theinformation/status/2071360628032184790