Italy’s antitrust and consumer rights regulator has opened a probe into Chinese AI start-up DeepSeek for allegedly failing to warn users with sufficient clarity about the danger of false information arising from so-called “hallucinations” in its chatbot.
The AGCM regulator said in a statement that DeepSeek did not give users “sufficiently clear, immediate and intelligible” warnings about the risk of misleading information in AI-generated content.
It said hallucinations were “situations in which, in response to a given input entered by a user, the AI model generates one or more outputs containing inaccurate, misleading or invented information”.
Data concerns
DeepSeek is under a separate investigation from Italy’s data protection regulator, which opened its probe in February while ordering the company to stop processing Italian users’ data and asking the government to block the app in the country.
The Garante data protection agency said at the time it had received “completely insufficient” answers to questions asked of the start-up.
“The limitation measure – adopted to protect the data of Italian users – follows the communication from the companies received today, the content of which was deemed completely insufficient,” the agency said at the time.
“Contrary to what the authority found, the companies declared that they do not operate in Italy and that European legislation does not apply to them.”
Italy’s Garante has been one of the EU’s most active regulators in policing AI models, in December 2024 fining OpenAI 15m euros (£12.6m) over the use of Italian users’ personal data, a fine the Silicon Valley firm said it would appeal.
The Garante temporarily banned OpenAI in Italy in March 2023 at the beginning of its investigation.
AI action
It also fined the company behind the Replika AI “virtual friend” 5m euros last month over an alleged failure to prevent children from using the service.
In February leading AI chip maker Nvidia filed a lawsuit against the European Commission for investigating its $700m acquisition of Israeli software company Run:ai, saying the Commission illegally went beyond restrictions on its ability probe small deals after accepting a referral request from Italy’s AGCM.
Nvidia said in a court filing that the Commission had accepted the referral request “based on the AGCM’s exercise of loosely defined, ex post, discretionary call-in powers”.