The European Unions privacy watchdog has stated that OpenAIs measures to reduce factually incorrect outputs from its ChatGPT chatbot are inadequate to fully comply with EU data regulations. According to a report released by the task force on Friday, while the efforts towards transparency are commendable, they fall short in ensuring data accuracy.Although the measures taken in order to comply with the transparency principle are beneficial to avoid misinterpretation of the output of ChatGPT, they are not sufficient to comply with the data accuracy principle, the task force emphasised.Concerns Raised by National RegulatorsThe task force, established by Europes national privacy watchdogs, was formed last year after national regulators, led by Italys authority, expressed concerns about the AI services widespread use. OpenAI did not immediately respond to Reuters request for comment regarding the report.Ongoing InvestigationsThe report noted that investigations by national privacy watchdogs in some EU member states are still underway. Consequently, a comprehensive description of the results is not yet available. The findings presented in the report should be viewed as a common denominator among the national authorities.Probabilistic Nature and Data AccuracyData accuracy is a fundamental principle of the EUs data protection regulations. The report highlighted the inherent limitations of the current training methods for AI models like ChatGPT, stating, Due to the probabilistic nature of the system, the current training approach leads to a model which may also produce biassed or made-up outputs.Furthermore, the report pointed out the potential risk of users accepting ChatGPTs outputs as factually accurate, regardless of their actual correctness. The outputs provided by ChatGPT are likely to be taken as factually accurate by end users, including information relating to individuals, regardless of their actual accuracy, the task force warned.
