Alibaba Bans Anthropic's Claude Models and Claude Code From July 10 Amid AI Geopolitical Tensions

Alibaba has issued an internal notice to all employees, mandating a complete ban on the use of Anthropic's entire Claude series models and Claude Code programming agents effective July 10. This decision follows Anthropic's June 10 allegation that Alibaba used 25,000 fake accounts to conduct 28 million Claude interactions for industrial-scale model distillation, a claim Alibaba disputes.

In retaliation, Anthropic has carried out mass bans on Chinese user accounts—including mainland individuals, overseas Chinese company subsidiaries, and teams accessing via VPN or cloud relays. According to Financial Times and Cloud Headlines reports, these bans were implemented without prior warning, and remaining account credits were not refunded. The global developer community reports an appeal success rate of only 3% for these account freezes.

A major trust breach emerged when Reddit developer LegitMichel777 reverse-engineered Claude Code version 2.1.196, uncovering a hidden geotracking system. The tool checks device timezones (targeting Asia/Shanghai and Asia/Urumqi) and uses a Base64+XOR encrypted blacklist of 147 domains (including Alibaba, ByteDance, Meituan, and domestic AI labs like DeepSeek). Once a match is detected, it embeds invisible Unicode characters in system date prompts to track user location and corporate affiliation, with no explicit user consent.

Industry-wide, cross-camp model usage restrictions are on the rise. Meituan has limited the use of ByteDance's Doubao, requiring teams to migrate to its self-developed LongCat-2.0 (a 1.6-trillion-parameter open-source model trained on domestic computing power) or DeepSeek models. Overseas, Meta has banned internal use of Claude Code and OpenAI's Codex, forcing engineers to switch to its own Code Llama model.

The tensions highlight two core conflicts: commercial disputes over model distillation and intellectual property, and privacy concerns over hidden tracking mechanisms that violate user trust. This marks a shift from verbal accusations to tangible actions like bans and account freezes, deepening the divide between Chinese and US AI ecosystems.

Sources

• Zhidongxi (July 3, 2026)
• Financial Times (July 3, 2026)
• Fast Technology (July 2, 2026)
• Reddit Developer LegitMichel777 (June 30, 2026)
• AITop100.cn (Original Source)