ChatGPT sued in Austria over his ‘hallucinations’

The Viennese privacy association NOYB announced on Monday that it has filed a lawsuit in Austria against the generative artificial intelligence program ChatGPT because it is “rampant” and even its publisher, OpenAI, “cannot stop it,” according to a statement cited by AFP and Agerpres.

These types of tools “regularly invent answers”, NOYB (acronym for “None of Your Business”, translated as “None of Your Business”) points out in the press release. The fact that AI systems based on pre-trained generative language models, such as ChatGPT, provide fictitious answers is a well-known fact, these false statements of the models being colloquially known as “hallucinations”.

The Viennese association notes that when it asked ChatGPT about the date of birth of NOYB founder Max Schrems, the chatbot “systematically provided false information” instead of admitting that it did not know. “If a system cannot provide fair and transparent results, it cannot be used to generate data about people. Technology must respect the law, not the other way around,” insists lawyer Maartje de Graaf, quoted in the press release.

This is simply “unacceptable”, according to NOYB, which points out that the European General Data Protection Regulation (GDPR) provides for the obligation of accuracy.

NOYB also states that the OpenAI firm “refused the plaintiff’s request to correct or delete the error, claiming it was impossible,” and that AI “unlearning” is actually a crucial issue surrounding the technology.

NOYB accuses the company behind ChatGPT of hiding its sources of information

The Viennese association also accuses OpenAI of not responding to the association’s request for access to its data and information sources, again violating the law. In its complaint, NOYB, which has become the bane of tech giants since its creation in 2018, asks the Austrian Data Protection Authority to open an investigation and fine the Californian company.

Launched in November 2022 and funded by Microsoft, ChatGPT has been taken by storm by users impressed by its ability to deliver essays, poems or translations in seconds. But, given the risks, procedures have been launched in several countries.

Italy temporarily blocked access to the ChatGPT website in March 2023 while the French regulator (CNIL) began to look into a number of complaints. A European working group was also established to encourage cooperation.

“It remains to be seen where this will take us,” said NOYB, which is currently skeptical of the authorities’ efforts to regulate artificial intelligence.