Italy cuts off ChatGPT due to privacy concerns

While ChatGPT has become what seems like a household name, the AI model’s method of data collection is somewhat concerning and has some clear negative connotations. With that being the case, Italy is moving forward with legal action to stop ChatGPT from operating for the time being.

As a large language model, ChatGPT bases most of its proverbial knowledge on one gigantic source – the internet. These models have to be trained over a period of time based on information accessible through the internet, which can be problematic, to say the least. Information is not always gathered through explicit consent, though said information is technically freely available on the web.

Today, Italy made those concerns known by taking action against OpenAI, essentially banning ChatGPT from running in the country for the time being. ChatGPT will receive a “temporary limitation on the processing of Italian users’ data.” Following this, Italy is launching an investigation to ensure their concerns are rightly placed. The main concern here is the likelihood OpenAI is breaching the European Union’s General Data Protection Regulation – GDPR.

What seems to have sparked the GDPR investigation into the AI language model was a security breach that took place on March 20. OpenAI indicated that both conversations submitted through ChatGPT from users as well as payment methods were potentially at risk in that breach.

By nature, ChatGPT aims to emulate human interaction. That interaction is executed so well on the AI’s part thanks to the complex training that OpenAI puts these models through. Unfortunately, that’s where there seems to be an overshadowing gray area. In that data collection, no Italian citizen is consenting to give their data to OpenAI. Even so, the company is able to collect it via freely accessible internet.

The Italian government thus notes that “there appears to be no legal basis underpinning the massive collection and processing of personal data in order to ‘train’ the algorithms on which the platform relies.” In essence, ChatGPT does not perform service or serve a larger purpose in Italy, since its main function is to act like a human and have conversations with people, as reductive as that sounds.

Obtaining data without explicit consent coupled with the fact that this data is often false makes for a strong case against the extremely advanced chatbot. Even still, an advanced GPT-4 model still relies on its trained data, and if the data it trains with is inherently false because it might have been pulled from any number of regular and social media sites, therein lies a massive problem.

The final concern voiced by Italy’s DPA is the lack of an age verification mechanism employed by OpenAI. With no restrictions, underage individuals have full access to whatever the chatbot throws at them. Though the service is meant for those over 13, the EU in particular has strong restrictions set in place, requiring these mechanisms to be active in some capacity.

OpenAI has undoubtedly done a lot of work to create something that not only impresses but blows up internationally. Unfortunately, if the AI has plans for reaching those in regions with stricter internet regulations, ChatGPT is probably going to have to play by the rules. Specifically, those set by the GDPR.

OpenAI has 20 days to adjust how ChatGpt operates in Italy and report those changes. Without those adjustments, the company is looking at a fine of up to 20 million euros or 4% of the total annual turnover of the company – whichever is less.