Italy’s antitrust authority closes probes into DeepSeek, Mistral, and Nova AI
The AGCM accepted binding commitments from all three chatbot providers, establishing a concrete benchmark for what ‘adequate’ hallucination transparency must look like in practice, and a 120-day compliance window before potential fines.
TL;DR
- Italy's AGCM closed investigations into DeepSeek, Mistral AI, and Nova AI after they agreed to commitments on AI hallucination warnings.
- The companies will implement more prominent and clear warnings about the potential for AI-generated inaccurate or misleading content.
- This action establishes a precedent for 'contextual' transparency, requiring warnings at the moment of risk.
- DeepSeek will also invest in reducing hallucination rates, a forward-looking obligation.
- Mistral AI will include in-chat disclaimers and strengthen its terms of service regarding output unreliability.
- Nova AI, an aggregator, must disclose its aggregator nature and that it doesn't process responses from underlying models.
- No fines were issued, but non-compliance within 120 days can lead to significant penalties.
- The AGCM's approach is expected to influence other EU regulators and the European Commission under the AI Act.