Según informes, ChatGPT acusó erróneamente a un hombre inocente de asesinar a sus hijos

It has been more than two years since ChatGPT made a significant impact on the global stage. While OpenAI has made several advancements, there are still a number of challenges that need to be addressed. One of the major issues is the occurrence of hallucinations, where false information is presented as factual. Recently, the Austrian advocacy group Noyb filed a second complaint against OpenAI for such instances of hallucinations. They highlighted a specific incident where ChatGPT inaccurately identified a Norwegian man as a murderer.

To make matters worse, when the man in question asked ChatGPT about himself, the AI claimed that he had been sentenced to 21 years in prison for the murder of two of his children and the attempted murder of a third. This hallucination was mixed with some real details, including the number of children he had, their genders, and the name of his hometown.

Noyb argues that such responses violate GDPR regulations. According to Noyb data protection lawyer Joakim Söderberg, personal data must be accurate, and users have the right to request corrections if needed. Simply providing a disclaimer that the chatbot can make mistakes is not sufficient when spreading false information.

Several other instances of ChatGPT’s hallucinations have been reported, including accusations of fraud, embezzlement, child abuse, and sexual harassment against individuals. These incidents have been documented by various publications.

Noyb’s first complaint to OpenAI, made in April 2024, focused on an inaccurate birthdate of a public figure. Despite the request to update or correct the information, OpenAI claimed that it could not alter data already in the system, only restrict its use in specific prompts. ChatGPT relies on a disclaimer stating that it «can make mistakes.»

While it is true that everyone makes mistakes, the question arises whether this logic applies to a widely used AI-powered chatbot. It remains to be seen how OpenAI will respond to Noyb’s latest complaint. This article was originally published on Engadget at .

FUENTE

Deja una respuesta

Tu dirección de correo electrónico no será publicada. Los campos obligatorios están marcados con *