Anthropic Blocks Accounts Linked to 5% to 7% of Revenue

Why did Anthropic block paying customers?

Anthropic alleges that Chinese AI developers used fraudulent accounts to extract Claude's capabilities and train competing models. The company banned tens of thousands of accounts as part of its crackdown on unauthorized model distillation.

Orbitrum Investor Impact

Why does losing this revenue matter for Anthropic?

Blocking these accounts removes a source of paid usage and helps explain Anthropic's slowing revenue growth ahead of its planned IPO. The company is accepting lower near term revenue to protect its technology and potentially limit competitors' access to its capabilities.

Has Anthropic permanently lost 5% to 7% of its revenue?

Not necessarily. Reuters reported that the blocked accounts represented 5% to 7% of revenue during the first nine months of 2026, not that total revenue subsequently declined by that percentage. The lasting financial impact depends on how much blocked usage would otherwise have continued and whether legitimate customer demand replaces it.

Have Anthropic's defenses stopped AI model copying?

Not completely. Anthropic estimates that its new safeguards have reduced the amount of Claude reasoning data extracted by unauthorized developers by approximately 90% to 99%. The company acknowledges that sophisticated attackers may develop new ways around these protections.

Could the crackdown affect Anthropic's IPO valuation?

Yes, through investors' assessment of sustainable revenue growth and competitive protection. Reuters reported that some investors were discussing an IPO valuation near $1.5 trillion, compared with earlier expectations around $2 trillion. The reporting does not establish how much of that change is attributable to the account bans.

What should Anthropic investors watch next?

The key test is whether legitimate customer revenue can sustain growth without the blocked accounts. Investors should also watch whether Anthropic maintains its reported reduction in unauthorized data extraction while avoiding significant disruption to legitimate developers.