Are OpenAI, Anthropic Overbilling You? Companies Claim $1.7 Mn In AI Overcharges

Companies deploying AI tools and models from OpenAI and Anthropic across their operations are being overbilled by an estimated $1.7 million in overcharges, according to a report by The Information.

Several major corporations, including Panasonic, HP, and Honda, are among those affected. Vaudit, a start-up specialising in auditing AI bills, reviewed $34 million worth of AI invoices from 60 companies between March and June 2026.

During the audit, the company found a major billing overcharge, with the majority of the billing issues traced to Anthropic's Claude Code, though invoices involving OpenAI services were also flagged.

Vaudit said it could not independently verify AI invoices because pricing depends on several factors, including the type of AI model, token usage, API calls, and others.

What May Have Caused Errors

The report attributed the overbilling to AI agents retrying tasks they were completing incorrectly, with customers unaware that the agent kept looping and accumulating charges.

Anthropic and OpenAI have both reportedly denied that billing errors are occurring, saying they have not seen evidence supporting the claims. Despite the denials, the report notes that companies are already spending massive amounts on generative AI models.

Since AI services are billed based on usage, such cost fluctuations could have a serious impact on how companies plan to rely on AI going forward.

Vaudit also found that after customers complained about inflated AI billing, providers such as Amazon, Google, Microsoft, Anthropic, and OpenAI reimbursed roughly 80% of the contested fees..

However, many companies have continued to use premium AI models even after the overcharges, while others are reported to be looking for cheaper alternatives. The episode has renewed calls for transparent and accurate billing practices amid growing AI compute costs.

Fight for Cheaper AI

Earlier this month, Microsoft introduced a range of low-cost AI models aimed at reducing expenses for customers driven by high AI costs. The company released Copilot Cowork, an independent AI system that lets users select from various models, including more affordable options.

Following this, reports suggested that leading AI firms OpenAI and Anthropic were planning significant price reductions for their artificial intelligence services, seeking to attract each other’s customers.

Microsoft CEO Satya Nadella had earlier revealed the company was considering deploying a version of DeepSeek, a China-based, ultra-low-cost AI provider that OpenAI and Anthropic have alleged copied and compressed their leading models.