Chinese Regulators Scrutinize DeepSeek and Moonshot AI After Anthropic Allegations

Beijing’s market supervision authorities have opened formal investigations into two homegrown generative‑AI firms, DeepSeek and Moonshot AI, after the U.S. company Anthropic reported that the Chinese startups were channeling user requests to its Claude model.

Anthropic, which develops the Claude family of language models, said internal logs showed a pattern of API calls originating from the two Chinese platforms that appeared to relay prompts and responses through Anthropic’s services. The company raised concerns that such routing could violate its licensing terms and potentially expose proprietary technology to foreign jurisdictions.

The probe comes amid a broader regulatory push in China that began in 2021, when the government introduced licensing requirements, data‑security reviews, and restrictions on the incorporation of overseas AI models into domestic products. Authorities have emphasized compliance with national security standards and the protection of intellectual property, and the current investigation is being framed as a routine enforcement of those rules.

Analysts note the timing is noteworthy, as diplomatic channels are reportedly preparing for a high‑profile meeting between U.S. President Donald Trump and Chinese President Xi Jinping. Some observers suggest the investigation could be positioned as a conciliatory gesture, signaling China’s willingness to address foreign concerns, while others see it as leverage in broader negotiations over technology trade and data governance.

If the inquiries confirm Anthropic’s claims, DeepSeek and Moonshot AI could face penalties ranging from fines to the suspension of their AI‑service licenses, and they may be required to redesign their systems to eliminate any unauthorized data flows. The case highlights the growing friction between rapid AI development and the evolving legal frameworks that govern cross‑border technology exchange, underscoring the need for clearer international standards as generative AI becomes increasingly global.