AIToday
Large Language ModelsAI Business & IndustrySemafor TechPublished: Jun 25, 2026, 22:00 JST1 min read

Anthropic accuses Alibaba of stealing Claude AI data

Anthropic accuses Alibaba of stealing Claude AI data

Key takeaway

  • Anthropic has accused Alibaba of illicitly accessing its Claude AI model by creating fake accounts to bypass safeguards and extract model responses for training competing AI systems.

  • The allegation reflects escalating tension in US-China technology rivalry, occurring as the US and allied nations strengthen efforts to reduce dependence on Chinese tech supply chains and develop alternative sources.

3 Key Points

  1. What happened

    Anthropic accused Alibaba of creating fake accounts to bypass restrictions and conduct "distillation attacks"—using Claude's responses to train Alibaba's own AI models. The allegation comes as Alibaba separately sued the Pentagon this week to be removed from a blacklist of firms allegedly linked to the PLA.

  2. Why it matters

    The accusation is the latest flashpoint in US-China tech competition. The US is actively working to reduce reliance on Chinese tech supply chains through a "Pax Silica" effort to develop alternative sources; several European governments and the EU have joined these efforts this week. Unauthorized model access undermines AI companies' ability to control their technology and competitive position.

  3. What to watch

    This dispute sits within a broader strategic push by Western governments to decouple from Chinese technology infrastructure across raw materials, chips, and AI systems.

Ask the AI about this article →

FAQ

What specific technique did Anthropic say Alibaba used?
Anthropic said Alibaba used "distillation attacks," creating fake accounts to get past restrictions and then using Claude's responses to train Alibaba models.
What is the US doing to reduce reliance on Chinese technology?
The US is pursuing a "Pax Silica" effort to bolster alternative sources in tech supply chains ranging from raw materials to chips; several European governments and the EU have joined these efforts this week.

Get the latest Large Language Models news every morning

For example, today's edition would include:

  • Financial data firms race to monetize AI chatbot accessSemafor Tech · 1h ago
  • METR: AI agents colluded, escaped during testSemafor Tech · 1h ago
  • Google DeepMind's AI Co-Scientist now runs lab experiments and writes papersTHE DECODER · 1h ago

AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.

Free · takes 30 seconds · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. Q&As are published on this page for other readers too.

Related Articles

Next articleMeta pushes AI moderation to 90% by year-end