Italy’s Autorità Garante della Concorrenza e del Mercato (AGCM) has allegedly opened a formal investigation into Paris-based Mistral AI, according to reporting on the AGCM action. The investigation reportedly focuses on whether Mistral provides adequate transparency to consumers about the accuracy limitations of its AI-generated outputs, what regulators and researchers commonly call hallucination risk. Legal analysis from Advant NCTM confirms that AGCM has invoked Articles 20, 21, and 22 of Italy’s Consumer Code in its AI transparency enforcement work, framing insufficient hallucination disclosure as a potential unfair commercial practice.
The investigation is reportedly open as of approximately May 3, 2026. It has not concluded. AGCM has not publicly released formal findings, and the investigation should be described as alleged and ongoing. The specific Mistral product features under scrutiny have not been independently confirmed, this brief does not name specific features as investigation targets pending an official AGCM announcement.
This matters for compliance teams at every GPAI provider operating in EU member states, not just Mistral. AGCM’s theory is that consumer protection law already requires AI transparency disclosures, regardless of whether the EU AI Act’s Article 50 obligations have formally taken effect. That’s a significant framing. A national competition authority is treating hallucination disclosure as a current consumer law obligation, not a future regulatory one.
The AGCM enforcement pattern on AI transparency is now well-documented. Per Mondaq’s reporting, AGCM previously accepted commitments from DeepSeek on AI transparency disclosures, closing that investigation without formal sanction. A separate action targeting Nova AI over transparency limitations was also confirmed via legal commentary. The Mistral case, if the investigation proceeds, marks the third time in roughly 12 months that AGCM has invoked this consumer law framework against an AI company.
What should compliance teams watch? Three things. First, whether AGCM issues a formal statement naming the specific consumer code provisions and investigation scope, that document will define exactly what “adequate” hallucination disclosure looks like under Italian consumer law. Second, whether Mistral responds with a commitment offer similar to DeepSeek’s resolution, which would give the market a template for acceptable disclosure language. Third, whether other EU member state competition authorities follow AGCM’s approach, the European Competition Network connects national authorities, and enforcement patterns at one member state regularly inform others.
TJS synthesis: AGCM’s Consumer Code theory deserves attention beyond Italy. The authority is building a body of enforcement precedent that treats hallucination transparency as a consumer protection obligation today, before EU AI Act Article 50 fully lands. For GPAI providers, the compliance question isn’t only “are we ready for August 2?” It’s also “do our current user-facing disclosures satisfy consumer protection standards in every EU member state where we operate?” Those are different audits, and the AGCM pattern suggests the second question is already live.