ChatGPT provides false information about people, and OpenAI can’t correct it

Under the GDPR (General Data Protection Regulation) in the EU, there are specific requirements regarding personal data of individuals:

• Accuracy: Information about individuals must be accurate.
• Access: Individuals have the right to access the information stored about them, including information about the source.
• Rectification: Individuals have the right to rectification under Article 16 GDPR if the data is inaccurate.
• Deletion: They can request that false information be deleted.
• Right to access: Under Article 15, companies must show which data they hold on individuals and the sources of that data.

OpenAI admits that it is unable to correct incorrect information on ChatGPT because the application generates responses by predicting the next most likely words in response to each prompt. This means that while OpenAI has extensive training data, there is currently no way to guarantee that ChatGPT is providing factually correct information. OpenAI refers to factual accuracy in large language models as an area of active research, indicating that the technology has not yet advanced to a point where it can ensure accuracy in every instance.

noyb, a data protection advocacy group, has filed a complaint against OpenAI with the Austrian Data Protection Authority (DPA). They are asking the DPA to investigate OpenAI's data processing practices, ensure the accuracy of personal data processed, comply with the complainant's access request, and bring its processing in line with the GDPR. Additionally, noyb requests that a fine be imposed to ensure future compliance.