On Thursday, France's privacy watchdog announced plans to interrogate DeepSeek to understand better the functioning of the Chinese startup's AI system and assess potential privacy risks for users. DeepSeek disclosed in a recent report that training DeepSeek-V3 required under $6 million in computing power from Nvidia H800 chips.
A spokesperson for the French watchdog stated, "The CNIL's AI department is currently analyzing this tool." They continued, "To comprehend the operations and data protection risks of this AI system, the CNIL will interview the provider of the DeepSeek chatbot."
This move by the French watchdog aligns with its active role in Europe, having fined tech giants like Google and Meta Platforms in the past. The authority had recently sought clarifications from DeepSeek regarding its handling of personal data. Meanwhile, Ireland's Data Protection Commission indicated it had received information from DeepSeek concerning data processing related to Irish users.
Within Europe, robust data protection measures are enforced through the General Data Protection Regulation (GDPR), known for its global reputation as a stringent privacy law. Violations of GDPR can result in fines of up to 4% of a company's worldwide revenue.
In efforts to regulate AI, the European Union enacted new regulations last year, imposing stringent transparency requirements on high-risk AI systems and less stringent obligations on general-purpose AI models. Penalties for non-compliance with the regulations range from 7.5 million euros (1.5% of turnover) to 35 million euros (7% of global turnover), depending on the severity of the violations.