DeepSeek, Mistral AI, and NOVA AI chatbots generated confident, well-structured answers to user queries without disclosing that their outputs could be inaccurate, hallucinated, or entirely fabricated — creating a risk of consumer harm in high-stakes areas like finance, health, and legal decisions. Italy's competition authority found that users had no way to distinguish correct answers from fabricated ones at the moment of interaction.
1 Answer
Expert: Italian Competition Authority (AGCM), Italy's consumer protection regulator On April 30, 2026, Italy's competition authority (AGCM) published the results of three separate investigations into DeepSeek, Mistral AI, and NOVA AI. The authority found that these AI chatbots could produce content that was incorrect, misleading, or entirely fabricated — without signalling that possibility to users at the moment of interaction. Acting under Articles 20, 21, and 22 of Italy's Consumer Code, the AGCM secured binding commitments from each company to warn Italian users explicitly, at the moment of use, that AI-generated content may be wrong. DeepSeek committed to the most extensive measures, including technical investment to reduce hallucinations across four categories: search-related inaccuracies, rewriting errors, legal hallucinations, and behavioural hallucinations. NOVA AI was also required to disclose that it provides unified access to multiple underlying AI models (OpenAI, Anthropic, Google, DeepSeek) without processing them itself. The AGCM noted that 'current technology cannot eliminate hallucinations entirely.' This is the first time a regulator has required AI companies to tell their users, at the moment of use, that fluency and accuracy are separate things. The companies have 120 days to demonstrate compliance or face fines up to €10 million.
Your answer
Sign in to verify this AI response.
Don't trust us — or the AI. Ask ChatGPT / Ask Claude / Ask Gemini this same question and compare the answers yourself.
More from this topic
While briefing cross-motions for summary judgment in a $12 million Oregon vineyard inheritance dispute, the plaintiff's attorneys filed three briefs that had been drafted with the help of an AI tool. The filings contained 15 references to "nonexistent cases" and eight fabricated quotations, all written in ordinary legal citation form, with case names, reporters and pin cites presented as if they were real authority. Nothing in the briefs told the court which passages came from the AI tool, or that the authority had not been checked. The fabricated material was only identified after the defendants' attorneys flagged it. Once the errors surfaced, attorney Stephen Brigandi deleted the false references and refiled the argument - without checking the rest of the briefs, which the court later found still contained misstatements of law. The court's opinion does not identify the specific AI product used, and no party produced the prompts or chat logs. The judge noted "circumstantial" but "persuasive" evidence that the client, Joanne Couvrette - described by the defense as a serial self-represented litigant - was likely the person who used AI to draft the briefs, but held that it was the attorneys who signed their names to the filings.
The family of a woman detained at the Adelanto ICE detention center in California went looking for an immigration attorney and found one on Facebook. It was an AI-generated avatar. The account told the family that a $4,355 bond payment would get their relative released from detention. They paid. The paperwork they submitted was never filed with any court - there was no attorney to file it. The same playbook runs in reverse against real lawyers. South Florida immigration attorney Angel Leal has been targeted by AI-generated videos that show him discussing immigration cases and telling people who need representation to message him on WhatsApp. More than 6,000 scams using his name and likeness have been found online.
Asked about Canadian fiddler Ashley MacIsaac, Google's AI Overview told searchers that he had been convicted of a series of criminal offences — sexual assault, internet luring, assaulting a woman and attempting to assault a minor — and that he was listed on the national sex offender registry for life. Every one of those statements was false. The summary was built from online articles about a different man in Atlantic Canada who shares his last name. MacIsaac found out in December 2025, when the Sipekne'katik First Nation cancelled a concert he was due to play that month, citing the AI summary, after members of the public complained about what they had read.