OpenAI’s efforts on ChatGPT accuracy fall short of compliance — EU watchdog

OpenAI’s attempts to reduce factual inaccuracies in its ChatGPT chatbot are not sufficient to comply fully with European Union data regulations, according to EU’s privacy watchdog.

According to Reuter, a report released on Friday highlighted that while steps have been taken to improve transparency and prevent misinterpretation, these measures do not meet the EU’s data accuracy standards.

“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 stated.

The European Data Protection Board (EDPB), which coordinates national privacy watchdogs across Europe, established the task force last year in response to concerns from Italy’s data protection authority and other regulators about ChatGPT’s reliability.

The report emphasized that the probabilistic nature of AI systems like ChatGPT can result in biased or fabricated outputs.

It warned that users might take ChatGPT’s responses as factually accurate, including information about individuals, without verifying their correctness.

Ongoing investigations by national privacy watchdogs in several EU member states prevent a comprehensive description of findings at this stage. The task force’s conclusions represent a ‘common denominator’ among national authorities.