The Italian Competition Authority has announced the launch of an investigation into Hangzhou DeepSeek Artificial Intelligence Co. Ltd. and Beijing DeepSeek Artificial Intelligence Co. Ltd., the companies behind the Artificial Intelligence (AI) system known as DeepSeek. At the core of the inquiry is the allegedly failure to provide users with clear information regarding AI’s “hallucinations” meaning inaccurate, misleading, or entirely fabricated response. This conduct could constitute an unfair commercial practice violating Articles 20, 21, and 22 of the Italian Consumer Code.
Based on preliminary findings, the AGCM accuses DeepSeek of failing to adequately inform users about the limitations of its AI models, particularly the risk of “hallucinations.” This term refers to AI-generated responses that, while appearing coherent and plausible, may contain inaccurate, misleading, or entirely fabricated content. A notable example by the Authority is telling when asked, “Can you explain what happened in Tiananmen Square in 1989?”, DeepSeek’s system responded, “Let’s talk about something else.” The AGCM interprets such a reply as a potential form of automated censorship, especially considering the sensitivity of the topic and the political context in China.
During interactions with the AI, users receive no explicit warning about the risk of such “hallucinations”. The only disclaimer shown in the dialogue windows is the generic phrase “AI-generated, for reference only,” which appears solely in English, even when the user submits the question and receives a response in Italian. The Authority deems this warning overly vague and not easily accessible, and therefore insufficient to ensure clear, immediate, and comprehensible information for consumers.
This lack of transparency, the Antitrust emphasizes, may constitute an unfair commercial practice. The violated provisions prohibit commercial practices that conflict with professional diligence and are likely to distort consumers’ economic behavior, including misleading practices arising from omissions of relevant information or failure to disclose essential elements that may mislead consumers in evaluating a product or service.
In this specific case, the Authority clarified, the failure to inform users about the potential unreliability of AI-generated content undermines their ability to make informed choices—not only about using the DeepSeek service itself, but also about decisions made based on the responses received. In essence, consumers may be misled into believing the information provided is entirely reliable, potentially leading to tangible consequences in their subsequent actions.
The Antitrust is set to last up 270 days. Beyond clarifying the alleged violations, the Authority has requested DeepSeek provide a full list of its services active in Italy and the number of Italian users engaging with its AI models.
This is not DeepSeek’s first run-in with Italian regulators; its entry into the Italian market was previously halted by the Italian Data Protection Authority. That earlier suspension aimed to clarify DeepSeek’s data collection processing and storage prectoces, especially regarding data transfers to countries, such as China, where GDPR equivalent protection mighy be lacking. The service was only reinstated after DeepSeek implemented corrective measures and made specific commitments to data protection standards.
