Anthropic Moves to Fully Block All Workarounds for Claude Access

According to a Financial Times report, Anthropic is launching an unprecedented enforcement campaign to eliminate all “underground” channels that allow users in restricted regions to access Claude. The policy, effective from July 3, 2026, explicitly bans personal overseas accounts, corporate‑provided Claude accounts, API relay stations, and even covert Azure‑based pathways.

The company has already deployed advanced detection mechanisms. Besides blocking suspicious IPs and dual‑currency credit cards, Anthropic’s system reads a user’s operating‑system timezone and other subtle signals to infer the user’s physical location. When a custom API forwarding address is set, the software checks the system timezone for prohibited regions.

Security researcher Adnane Khan discovered that Claude Code version 2.1.91 (released April 2026) embeds hidden logic: each request includes a system prompt containing a date string like “Today's date is 2026‑06‑30”. If the request matches a restricted region, the code replaces hyphens with slashes and swaps the apostrophe in “Today's” with visually identical Unicode characters, effectively masking the detection trigger.

Anthropic previously warned in September 2025 that it would broaden restrictions to cover overseas subsidiaries of companies attempting to circumvent the policy. The July 2026 rollout intensifies this stance, allocating substantial resources to locate and shut down relay stations, which the company claims are rarely used by top AI firms due to concerns over exposing proprietary code.

Examples of targeted workarounds include:

Employees using personal overseas accounts funded by their employer.

Companies providing enterprise‑level Claude accounts to staff.

Global enterprises leveraging compliant Singapore subsidiaries to purchase Claude via Azure, then routing results back through internal networks to domestic engineers.

API relay services that forward requests to Claude and return responses.

All of these are now classified as violations, and affected accounts have been disabled without refunds or meaningful appeal processes.

Following the crackdown, Alibaba announced an internal ban on all Anthropic products, including Sonnet, Opus, Fable, and Claude Code, effective July 10. The article ends by questioning whether domestic models such as GLM, DeepSeek, Qwen, and Step will gain market share as a result.