Italy closes antitrust probes into DeepSeek, Mistral, Nova AI over AI hallucinations

Italy’s competition and consumer watchdog, the AGCM, has officially closed its antitrust investigations into three major AI chatbot providers: China’s DeepSeek, France’s Mistral AI, and Nova AI. The probes, which centered on the issue of AI hallucinations , instances where chatbots generate false or misleading information , have been resolved after all three companies agreed to binding commitments.

The AGCM accepted these commitments, which establish a concrete regulatory benchmark for what constitutes “adequate” transparency around hallucination risks. This marks a significant step in defining accountability standards for generative AI systems operating in the European market. Under the terms of the agreement, each provider must now comply with a 120-day implementation window to adjust their practices. Failure to meet these requirements could result in financial penalties.

By closing the cases with binding commitments rather than fines, the Italian authority has effectively created a practical framework for how AI companies should handle the disclosure of potential inaccuracies. This move signals a shift toward proactive regulation, where companies are given the opportunity to self-correct before facing sanctions. The outcome also provides clarity for other AI developers operating in Europe, as it sets a precedent for what transparency obligations look like in the rapidly evolving chatbot landscape.