Key Takeaways
- Alibaba will prohibit its employees from using Anthropic’s Claude AI tools for work starting July 10, citing alleged back‑door security risks.
- The ban follows Anthropic’s accusation that Alibaba carried out “the largest known distillation attack” on its models and its request that Chinese firms be barred from accessing its technology.
- Alibaba staff must uninstall all Claude‑related products and switch to the company’s internal AI assistant, Qoder.
- Anthropic’s terms of service already restrict usage by Chinese entities and other “adversarial nations,” a policy reinforced after revelations that Chinese firms accessed Claude via third‑country routes.
- Reports indicate that Ant Group provided corporate Claude accounts through its Singapore‑based intranet, while ByteDance launched a reimbursement program allowing engineers to expense personal Claude subscriptions accessed via VPNs.
- Both Alibaba and Anthropic declined to comment on the developments; the situation reflects widening tensions over AI technology transfer between the U.S. and China.
Alibaba’s Internal Ban on Anthropic AI
Alibaba has instructed its workforce to cease using any Anthropic artificial‑intelligence products for professional tasks effective July 10. According to individuals familiar with the internal directive who requested anonymity, the Chinese e‑commerce conglomerate placed Anthropic’s Claude Code on a “high‑risk software list.” The move is justified by concerns that the U.S.‑based AI firm may possess undisclosed back‑door vulnerabilities that could compromise Alibaba’s data security. As one source told CNBC, “Alibaba employees are required to uninstall all Anthropic models and agent products and instead use the Chinese company’s own AI assistant, Qoder.” The directive leaves little room for interpretation: any continued reliance on Claude for work‑related activities will be considered a policy violation.
Anthropic’s Accusations and Legal Stance
The ban comes on the heels of a sharply worded letter Anthropic sent to the U.S. Senate Committee on Banking, Housing, and Urban Affairs in June. In that correspondence, the AI startup alleged that Alibaba was “brazenly” and “illicitly” attempting to siphon its proprietary AI capabilities. Anthropic went further, claiming that Alibaba had executed “the largest known distillation attack” on its models to date—a technique where a smaller model is trained to mimic the behavior of a larger, more sophisticated one, effectively extracting its knowledge without direct access to the underlying weights. Anthropic’s terms of service explicitly prohibit Chinese companies and entities from “adversarial nations” from using its models, a clause the firm says it is now enforcing more vigorously after uncovering circumvention attempts.
Work‑Arounds and Third‑Country Access
Investigations by the Financial Times revealed that despite Anthropic’s restrictions, several Chinese firms had found ways to access Claude through intermediary jurisdictions. Ant Group, the financial technology affiliate of Alibaba, reportedly supplied employees with corporate Claude accounts that were reachable via its intranet, which is linked to a Singapore‑based entity. ByteDance, the parent of TikTok, did not facilitate direct access but introduced a reimbursement program in April that permits engineers to expense personal Claude subscriptions. Those subscriptions could then be utilized through virtual private networks (VPNs), effectively bypassing geographic blocks. A source familiar with ByteDance’s policy told CNBC that the initiative aims to “encourage staffers to experience and learn about a wider range of AI products to enhance their skills.” Both Ant and ByteDance declined to comment on the FT’s findings when approached by CNBC.
Corporate Responses and Ongoing Tensions
When approached for comment, neither Alibaba nor Anthropic provided statements. Alibaba’s silence on the matter leaves its internal risk assessment undisclosed, while Anthropic’s refusal to elaborate suggests it may be preferring to let its legal and compliance channels address the issue publicly. The episode underscores a broader trend of heightened scrutiny over AI technology transfers between the United States and China. As governments and corporations alike grapple with the strategic implications of generative AI, incidents like this highlight the tension between fostering innovation and safeguarding national security interests. The reliance on VPNs and third‑country routing illustrates how determined actors can navigate existing restrictions, prompting policymakers to consider more robust enforcement mechanisms.
Implications for the AI Landscape
Alibaba’s decision to replace Claude with its homegrown Qoder assistant signals a push toward self‑sufficiency in AI capabilities, a goal echoed by many Chinese tech giants seeking to reduce dependence on foreign models. For Anthropic, the episode may accelerate efforts to close loopholes that allow circumvention of its usage policies, potentially leading to stricter verification processes or geo‑fencing technologies. Meanwhile, companies such as Ant and ByteDance may reassess their internal AI training programs, balancing the desire to expose employees to cutting‑edge tools with the need to comply with evolving international regulations. As the global AI race intensifies, the interplay between corporate policy, legal frameworks, and technological workarounds will continue to shape how firms access—and protect—advanced artificial‑intelligence resources.
https://www.cnbc.com/2026/07/06/alibaba-anthropic-ai-ban-claude-china.html