Anthropic's prospectus says revenue from 'violating users' made up about 5%–7% of its total revenue in the first nine months of this year.

Anthropic disclosed in its prospectus that customers found to violate its usage policies or commercial terms contributed 5% to 7% of its revenue in the first nine months of the year; Reuters reported the company cut off some paid accounts and accused some of them of links to Chinese AI labs that obtained Claude outputs to train their own models. Anthropic says it sharply reduced this revenue in the third quarter through stronger defenses, cutting distillation extraction by 90% to 99% and banning tens of thousands of accounts. The disclosure matters because it puts a first number on how much model distillation costs a leading AI company, and on the defensive price of IP and national-security pressure.

• LinuxDo2026-10-26 23:30