What happened
The Financial Times reported that Anthropic is actively detecting and blocking unauthorized Claude access by Chinese firms. The company documented patterns such as a corporate account routed through a Singapore subsidiary and VPN-based personal subscriptions that engineers were reimbursed for.
Anthropic frames the activity as distillation attacks intended to extract Claude's capabilities. Its new detection includes monitoring computer time zones and targeting overseas relay services.
Why it matters
The move marks a shift from passively enforcing terms of service to actively detecting and blocking evasion. That is a meaningful change in how a frontier lab polices who uses its models.
It also reflects how export and access controls are pushing labs to build enforcement into their products, not just their contracts.
MintedBrain take
Stricter detection can catch legitimate users who happen to work across borders or behind a VPN. If your team spans regions, keep access arrangements clean and documented so routine setups are not mistaken for evasion.
Discussion
Sign in to comment. Your account must be at least 1 day old.