A new report has surfaced from the ChatGPT taskforce, initiated by the European Data Protection Board (EDPB), shedding light on OpenAI’s efforts to address the risk of the AI generating factually inaccurate content, which allegedly don’t fully comply with EU data regulations.
The EDPB, while acknowledging measures taken by OpenAI to uphold transparency principles aiming to prevent misinterpretation of ChatGPT’s output, has raised concerns about the adequacy of these measures in ensuring data accuracy as mandated by EU regulations.
“As a matter of fact, due to the probabilistic nature of the system, the current training approach leads to a model which may also produce biased or made up outputs”,
“In addition, 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 report said.
Highlighting the paramount importance of accuracy in EU data protection laws, the report expressed worries about ChatGPT’s probabilistic nature potentially leading to biased or fictitious outputs.
Also, despite its probabilistic nature, end users are prone to perceive outputs as factually accurate, regardless of their actual accuracy.
Released on May 24th, the report dictates that OpenAI must implement suitable measures both in determining processing methods and during actual processing to align with data protection policies, adhere to GDPR requirements, and safeguard data subject rights.
The report stressed that GDPR compliance should not be delegated to data subjects through terms and conditions clauses.
It also emphasised the taskforce’s role in facilitating information exchange among supervisory authorities (SAs) regarding interactions with OpenAI, enabling enforcement actions concerning ChatGPT, and identifying areas necessitating a unified approach across various enforcement actions by SAs.
“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 said in a report released on its website.
While investigations are ongoing, a comprehensive report is yet to be issued. Nonetheless, the stances presented in the report indicate a consensus among SAs on the interpretation of GDPR provisions regarding matters under investigation.
