Anthropic accuses Alibaba of AI model theft via massive API abuse (heise.de)

• Nearly 29 million API calls from ~25,000 fake accounts to Claude over six weeks.
• Method used is model distillation, prohibited by terms of service.
• Anthropic urges US Congress to block Chinese access to AI chips and allow firms to share threat data.
• Elon Musk admitted using distillation from OpenAI for Grok.
• Anthropic's Fable 5 model was temporarily blocked by US government.

"Anthropic alleges Alibaba-linked attackers used nearly 25,000 fraudulent accounts to access its Claude AI model about 29 million times between late April and early June 2026, aiming to copy its logic through model distillation. Anthropic urges US Congress to tighten restrictions on Chinese firms."