Australian Financial Review
Details
- Date Published
- 25 June 2026
- Priority Score
- 4
- Australian
- Yes
- Created
- 25 June 2026, 02:00 am
Authors (1)
Description
The AI company says Chinese e-commerce giant Alibaba used swarms of fake accounts to illicitly “extract” the chatbot’s capabilities.
Summary
Anthropic has alleged that Alibaba's Qwen AI lab conducted a systematic campaign to 'extract' technical capabilities from the Claude model through industrial-scale fraudulent account usage. This incident highlights critical vulnerabilities in frontier AI security and model theft, specifically regarding the unauthorized transfer of advanced agentic reasoning and software engineering capabilities to geopolitical rivals. Such illicit access undermines global safety export controls and poses significant risks to the competitive advantage and safe deployment of high-capability AI systems. The situation underscores the urgent need for robust international governance frameworks to prevent the proliferation of frontier AI models that bypass localized safety guardrails and regulatory oversight.