Italy first country blocks DeepSeek AI over data privacy concerns

MediaNama reports, citing Reuters, that the Chinese AI model DeepSeek R1 has been blocked by Italy, the first country to do so, because too little is known about how it handles users' data. The article says Italy's data protection regulator, Garante Privacy, confirmed the restriction in a tweet and said it had at the same time opened an investigation into the model.

According to MediaNama, the regulator had sent DeepSeek a request for information on January 28, asking the company to say which personal data the model collects, where that data comes from, why it is gathered, what the legal basis is, and whether the storing servers are in China. The regulator also asked about training data and, if it came from scraping the web, how registered and unregistered users were told about the processing of their data.

The article describes several wider concerns about the model, which it says was released on January 20 and has been controversial from the start. It says many people claim the model heavily censors answers that may conflict with the Chinese government, and that OpenAI is reported to have said DeepSeek used its API to build on OpenAI's models. MediaNama also cites a blog post by the cloud security firm Wiz, which found a publicly accessible database exposing chat history, secret keys and backend details.

The piece closes by wondering whether Italy will be the first of several countries to act, and notes that the US Navy has told its personnel not to use DeepSeek.