An independent audit by billing-verification startup Vaudit reviewed $34 million in AI invoices across 60 companies between March and June 2026 and found roughly $1.7 million in overcharges tied to Anthropic and OpenAI services. Errors included customers being billed for newer, more expensive models when cheaper ones were actually used, and charges applied even when agents failed to complete a task or returned an error. After Vaudit flagged the discrepancies, Amazon, Google, Microsoft, Anthropic, and OpenAI credited back about 80% of the disputed charges, though Anthropic and OpenAI disputed the characterization and said they saw no evidence of widespread overbilling.