Italy’s data privacy regulator has committed to lifting its ChatGPT ban, potentially as early as the end of April, if OpenAI makes prescribed changes to bring it into compliance with EU regulations.
OpenAI can get out of the ChatGPT ban if it improves transparency, allows users to view and delete stored data, and implements added safeguards to protect minors. April 30 is the earliest point at which the service could return to Italy, but the regulator says that the terms must be met by September 30 at the latest or the ban becomes more permanent.
ChatGPT ban reversal requires numerous changes, and a nation-wide ad campaign
To get out from under the ChatGPT ban, OpenAI will need to age-gate the service to ensure that users are at least 18, collect explicit consent from users to add their data to its training models, and allow anyone (whether they have used the service or not) to check and see if the system is holding false personal information and to request deletion of it.
OpenAI will also have to post a privacy notice on its website describing how it stores and uses collected data, and will have to run an advertising campaign across Italy on “radio, television, newspapers and the web” to notify the population of all of these changes, according to the Guarantor for the Protection of Personal Data (GPDP).
The changes are meant to address an assortment of privacy concerns that led to the ChatGPT ban. Foremost among them was that minors were entering sensitive personal information without required guardrails and oversight. The GPDP was also critical of ChatGPT’s transparency into data collection and use practices, and that users had no means to see if the system was holding their protected personal information or making use of it in some way.
In terms of age-gating, the GPDP is requiring that children under 13 be banned entirely from using the service. Those age 13 to 18 must provide the consent of a parent or guardian when creating an account. The order calls for an “age verification process” but it is not clear what this will entail beyond self-certifying one’s age. A recent blog post from OpenAI indicates that the firm is “looking into” options in this area.
ChatGPT bans under consideration in multiple countries, but are counterbalanced by competitive concerns
Italy is thus far the only country in western Europe to issue a ChatGPT ban, but it has reportedly also been considered by Germany (who made contact with Italy about their ban) and is now facing an investigation from France’s lead data protection agency after a number of privacy complaints. Spain has asked the European Data Protection Board to introduce the topic of chatbot regulation to nations across Europe in the interest of developing unified standards.
But outside of Canada, which has opened a probe similar to the investigation in France, the idea of cracking down on ChatGPT still seems to be off the table. The only other countries that have banned ChatGPT are generally those that OpenAI will not allow to use it (such as China, North Korea and Russia), and some of these nations have gone full steam ahead with their own comparable large language models. There is some fear of an “AI arms race” as governments ponder the national security and propaganda implications, not to mention the boosts it could provide to private sector productivity and innovation, and many governments appear to be concerned about getting left behind if they slow things down too much in the name of privacy and safety.
Regulations are almost inevitable, however, given the potential harms that this relatively immature version of ChatGPT has already demonstrated. The overall impact on economies remains unpredictable, as industries of all types look to replace human workers with a more inexpensive technological solution. This transition may already be happening too fast for the present level of ChatGPT’s capability, however, as the chatbot has a far from perfect track record in its output. There is also serious concern about the impact of such tools on education and general information literacy, as students attempt to get AI to do homework and take tests for them.
Dr. Ilia Kolochenko, Founder of ImmuniWeb, sees regulation taking off in the near future in the same way the activation of the GDPR in Europe prompted a response around the world: “Privacy issues are just a small fraction of regulatory troubles that generative AI, such as ChatGPT, may face in the near future. Many countries are actively working on new legislation for all kinds of AI technologies, aiming at ensuring non-discrimination, explainability, transparency and fairness – whatever these inspiring words may mean in a specific context, such as healthcare, insurance or employment. Of note, the regulatory trend is not a prerogative of European regulators.”
“For example, in the United States, the FTC is poised to actively shape the future of AI. The Cyberspace Administration of China is also energetically working on new rules and restrictions for AI companies. One of the biggest issues is training data, which is frequently collected and used by AI vendors without any permission from content creators. While modern intellectual property (IP) law provides from little to no protection to copyrighted content, most large-scale data-scrapping practices likely violate terms of service of digital resources, such as online libraries and websites, and may eventually lead to an avalanche of litigation for breach of contract and interrelated claims. Some jurisdictions may even wish to criminally prosecute such practices under their unfair competition laws. That being said, banning AI is a pretty bad idea: while law-abiding companies will submissively follow the ban, hostile nation-state and threat actors will readily continue their research and development, gaining unfair advantage in the global AI race,” added Kolochenko.
Italy has been quicker to take these tools offline than other nations, having previously banned the companionship-focused chatbot Replika.ai in early February. There were similar concerns about how minors might be impacted with the chatbot proving quick to move to sexualized responses, and what personal information minors might end up putting into the system. Chatbot bans are not an entirely new phenomenon, however; in 2019 Facebook twice had to ban a chatbot operated by Israel Prime Minister Benjamin Netanyahu’s page for taking it upon itself to illegally share polling information and promote conspiracy theories about Arab residents.