Anthropic, OpenAI customers overbilled by $1.7 million due to billing errors

In a startling revelation, numerous companies utilizing AI tools and models from OpenAI and Anthropic have faced significant billing inaccuracies. An audit conducted by the startup Vaudit uncovered that these businesses have been overcharged by approximately $1.7 million due to errors in invoicing. The audit, which scrutinized $34 million worth of AI bills from 60 different firms between March and June 2026, highlighted that most discrepancies stemmed from services related to Anthropic's Claude Code, though issues were also noted with OpenAI's offerings. Laura Bratton of The Information provided insight into the potential causes behind these billing errors. She suggested that certain AI agents may have been repeatedly attempting tasks inaccurately, resulting in customers being unaware of ongoing charges for failed attempts. Despite these findings, both Anthropic and OpenAI have refuted claims of widespread billing errors, stating that they have not encountered evidence supporting such allegations. This revelation comes at a time when companies are heavily investing in generative AI models, which are billed based on usage. The fluctuating costs can significantly affect corporate budgeting and reliance on AI technology. In response to complaints regarding overbilling, major AI service providers, including Amazon, Google, Microsoft, Anthropic, and OpenAI, took steps to rectify the situation by refunding approximately 80% of disputed charges. Despite the billing concerns, many companies continue to leverage premium AI models, although there is a growing trend of exploring more affordable alternatives. This situation underscores the critical need for transparent and accurate billing practices in the rapidly expanding AI sector.