AI company Anthropic is stepping up efforts to detect and shut down unauthorized use of Claude from China routed through cloud providers and overseas subsidiaries, according to a report published on July 3, 2026. The company is targeting workarounds—proxies, VPNs, fake accounts and foreign entities—to make its geographic restrictions more effective, as reported .
Sept 4, 2025 · Anthropic
Anthropic Cracks Down on China's Backdoors into Claude
The company is detecting and shutting down workarounds — via cloud providers and overseas subsidiaries — that let major Chinese firms including Ant Financial and ByteDance access its coding model Claude Code despite an explicit ban.
~25,000
alleged fake accounts tied to an Alibaba-affiliated operator (June 2026)
28.8M+
interactions run through those unauthorized accounts
~30
organizations hit by an AI-led espionage campaign abusing Claude Code (Sept 2025)
Four main loopholes — and the response
How access flowed around a ban on PRC use and PRC-controlled subsidiaries.
Via overseas subsidiary
Corporate accounts set up through a Singapore-based entity, used on internal intranet.
VPN + personal plan
Employees given individual subscriptions over VPN, with costs reimbursed.
Cloud API
API access from overseas subsidiaries via Microsoft Azure and similar.
Intermediary service
"Transfer station" relays route mainland traffic to overseas accounts.
Why the crackdown
National-security risk of use by military and intelligence agencies
Concern over "distillation" — harvesting US model outputs to train Chinese models
Compute access that circumvents export controls
User verification + blocked Chinese bank payments (Microsoft supports enforcement)
Why it keeps happening
Claude Code's coding ability is rated above domestic Chinese models
Strong incentive for engineers and startups to find workarounds
A Claude Code tracking feature was partly rolled back over privacy concerns
Rivals seen as easier to reach via VPN and weaker on verification
The enforcement loop
Detect workaroundtime-zone & usage checks
→
Block accountsverify users, cut payments
→
Expand bannow covers subsidiaries
→
Update toolsongoing detection
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…