Stock Markets July 3, 2026 12:44 AM

Anthropic ramps up defenses after Chinese firms find routes to Claude

Company detects access via overseas subsidiaries, cloud platforms and VPN-enabled personal accounts and is intensifying monitoring and takedowns

By Maya Rios
Share
Twitter Reddit Facebook LinkedIn
MSFT

Anthropic has increased efforts to prevent unauthorised access to its Claude AI models from Chinese companies after identifying multiple workaround methods. The company found use of overseas subsidiaries, cloud infrastructure and virtual private networks to reach its services, practices that it says violate its terms of service even if they do not breach U.S. or Chinese law.

Anthropic ramps up defenses after Chinese firms find routes to Claude
MSFT
Summarize with
ChatGPT Perplexity Claude Grok Gemini

Key Points

  • Anthropic identified multiple workaround methods used by Chinese-linked firms to reach its Claude AI models, including overseas subsidiaries, cloud providers and VPN-enabled personal accounts - impacting AI service providers and cloud infrastructure sectors.
  • Ant Financial reportedly provided corporate Claude accounts tied to a Singapore entity, while ByteDance reimbursed engineers for personal subscriptions accessed via VPNs - highlighting corporate compliance and employee access practices.
  • Anthropic is increasing detection and takedown efforts, monitoring account indicators such as computer time zones and targeting "transfer station" services that relay requests through overseas Claude accounts.

Anthropic has begun tightening controls aimed at blocking access to its Claude artificial intelligence services by entities linked to China, after internal checks uncovered a range of bypass methods used by some companies, according to people familiar with the matter.

Sources who spoke with reporters said several firms have been able to reach Claude through routes that skirt Anthropic's restrictions. That access has included the use of overseas subsidiaries, third-party cloud providers and other technical workarounds.

Two specific examples cited are Ant Financial and ByteDance. In the case of Ant Financial, employees were given corporate Claude accounts that were registered under a Singapore-based entity. ByteDance reportedly reimbursed individual engineers for personal Claude subscriptions that were accessed while connected to virtual private networks, or VPNs.

The practices identified do not appear to violate either U.S. or Chinese law, the sources said, but they are in breach of Anthropic's terms of service. Those terms specifically bar Chinese companies and foreign entities under their control from using Anthropic's models.

In response, Anthropic has intensified technical measures to detect and disable such access. Measures described include monitoring account activity for telltale indicators - for example, analysing users' computer time zones - and taking aim at so-called "transfer station" services that relay requests through overseas Claude accounts.

The reporting also notes that some companies access Claude by routing traffic through foreign subsidiaries that run on cloud infrastructure. Microsoft Azure was given as an example of the cloud platform used to facilitate such access.

Anthropic's actions centre on identifying patterns in account behaviour and cutting off intermediary channels that allow requests to be proxied through accounts based outside China. The company appears to be pursuing a mix of account surveillance and targeted takedowns to uphold its contractual restrictions.


What this covers

  • Which companies were named: Ant Financial and ByteDance were cited as having used overseas corporate accounts and VPN-enabled personal subscriptions respectively.
  • Methods of access: overseas subsidiaries, cloud providers and transfer services relaying requests through foreign Claude accounts.
  • Enforcement approach: anthropic is monitoring account signals such as time zone data and targeting intermediary services that reroute requests.

Risks

  • Enforcement risk: Anthropic's terms of service prohibit access by Chinese companies and controlled foreign entities, but detecting and blocking proxying through subsidiaries and relays is technically challenging - this affects AI service providers and enterprise compliance teams.
  • Operational uncertainty: Use of cloud infrastructure by overseas subsidiaries, including platforms such as Microsoft's Azure, can complicate identification of controlled access and may require ongoing technical monitoring by cloud and AI operators.
  • Policy limitation: Practices described reportedly do not violate U.S. or Chinese law even as they breach Anthropic's terms, creating a gap between legal permissibility and contractual enforcement for technology and legal teams.

More from Stock Markets

Kuaishou Rally Follows Record RMB 19 Billion Raise for Kling AI Unit Jul 3, 2026 Samsung Electronics Rockets Higher After Steep Sell-Off, Driven by AI Optimism and Bargain Buying Jul 3, 2026 AUSTRAC signs off on Sportsbet's anti-money laundering fixes after remediation undertaking Jul 3, 2026 Tesla's Chinese Suppliers Jump After Strong Q2 Deliveries; Market Optimism Rises Jul 3, 2026 Sumitomo Chemical Shares Jump After Korean Unit Agrees JV With Samsung Electro-Mechanics Jul 3, 2026