Key Takeaways (TL;DR): Anthropic may have earned up to $2.3 billion from users who violated Claude's terms of service over nine months in 2026. These rule-breakers—often deploying the AI for banned activities like automated content generation or reverse engineering—accounted for 5–7% of the company's revenue. For privacy-conscious users, a free auto-renewing subscription is available at /sub/happ/for-russia, and a private anti-DPI multi-key with white-list bypass is accessible via the Telegram bot @yavpn_robot.
According to a prospectus filed ahead of its IPO, Anthropic's AI assistant Claude generated a surprising chunk of revenue from users who were actively breaking the rules. Vedomosti calculated that these violators contributed between $1.66 billion and $2.33 billion—or 5–7% of total revenue—over the first nine months of 2026. The figure highlights a growing tension in the AI industry: as companies tighten usage policies, they also rely on the very users they ban.
Who are these rule-breakers?
Anthropic's terms of service prohibit using Claude for activities like mass content generation, automated decision-making in high-risk domains, reverse engineering, or training competing models. Yet a subset of customers—from startups to individual developers—routinely ignore these restrictions. The company's own data suggests that detecting and blocking such usage is a cat-and-mouse game, with violators often masking their activities through API key rotation or proxy chains.
Why the revenue matters
For a company preparing to go public, every revenue stream counts. The $1.66–2.33 billion range is not trivial—it could cover Anthropic's compute costs for months. But it also poses a reputational risk: if investors learn that a significant portion of income comes from users the company officially disavows, it could raise questions about enforcement and governance. Anthropic has stated it is investing in better detection systems, but the financial incentive to look the other way is clear.
What this means for users
For ordinary users, the takeaway is twofold. First, AI companies are watching—and they're getting better at spotting violations. Second, if you're using Claude for legitimate purposes, you're effectively subsidizing the rule-breakers who drive up infrastructure costs. If you're concerned about privacy or want to avoid being lumped in with abusers, consider using a dedicated proxy or VPN to keep your traffic separate. A free auto-renewing subscription is available at /sub/happ/for-russia, and a private anti-DPI multi-key with white-list bypass is accessible via the Telegram bot @yavpn_robot.