Anthropic, the creator of the Claude series of large language models, has lodged a formal complaint accusing Alibaba of illicitly extracting core model capabilities. The allegation surfaced after Anthropic detected unauthorized access patterns that mirrored internal testing data. By claiming theft of proprietary AI techniques, Anthropic signals a rare clash between a U.S. AI startup and a Chinese cloud giant publicly today.

Alibaba’s cloud division, which offers AI inference services to enterprise clients, reportedly integrated Claude‑derived algorithms into its own offerings without permission. Engineers allegedly reverse‑engineered the model by capturing API responses and reconstructing the underlying architecture. If proven, the breach could significantly expose Alibaba to intellectual‑property lawsuits and force a reevaluation of cross‑border data‑sharing protocols in the fast‑moving generative‑AI market now.

Anthropic plans to pursue legal remedies and has urged regulators to scrutinize Alibaba’s data‑handling practices. The dispute arrives as both firms race to monetize Claude’s advanced reasoning and coding abilities, which competitors consider among the most sophisticated outside of OpenAI’s suite. A court ruling could set a significant precedent for how AI model internals are protected under international copyright law globally.