By Foo Yun Chee
BRUSSELS (Reuters) – France’s privacy watchdog said on Thursday it will question DeepSeek to gain a better idea of how the Chinese startup’s AI system works and any possible privacy risks for users.
DeepSeek attracted global attention after writing in a paper last month that the training of DeepSeek-V3 required less than $6 million worth of computing power from Nvidia H800 chips.
“The CNIL’s AI department is currently analysing this tool,” a spokesperson for the French watchdog said.
“In order to better understand how this AI system works and the risks in terms of data protection, the CNIL will question the company that offers the DeepSeek chatbot,” they added.
The French watchdog is one of the most active in Europe and has fined Google and Meta Platforms, among others.
Italy’s data protection authority said earlier this week it was seeking answers from DeepSeek on its use of personal data, while Ireland’s Data Protection Commission said it has requested information from the Chinese company about data processing conducted in relation to Irish users.
Europe has been particularly protective of privacy rights and its General Data Protection Regulation (GDPR) is widely regarded as one of the most comprehensive and stringent data privacy laws in the world.
GDPR breaches can lead to fines up to 4% of a company’s total global turnover.
The 27-country European Union last year agreed to landmark rules which impose strict transparency obligations on high-risk AI systems and lighter ones for general-purpose AI models.
Fines for violations range from 7.5 million euros, or 1.5% of turnover, to 35 million euros or 7% of global turnover depending on the type of violations.
(Reporting by Foo Yun Chee; Editing by David Goodman and Alexander Smith)
