The European Data Protection Board (EDPB)’s ChatGPT taskforce has released another report pointing out the limitations of the generative artificial intelligence (GenAI) tool, as it once again fails to pass European data accuracy standards.
The report points out that ChatGPT’s current training revolves around probability-based text generation, leading to biased or made-up outputs.
“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 reads.
Data accuracy is one of the core principles of data protection regulations in the European Union.
ChatGPT is also under fire for breaching data privacy rules in multiple countries, with investigations or outright bans launched in Canada, the US, Italy, Germany, Spain, China and France.
ChatGPT must admit to inaccuracies
The report insists that generative AI such as ChatGPT are required under Article 5 of the General Data Protection Regulation to be transparent in the accuracy of the information given to users.
The taskforce asserts that clear information on the accuracy of the generated information is required to give users a meaningful understanding of the information’s value, “including explicit reference to the fact that the generated text, although syntactically correct, may be biased or made up”.
The investigation is ongoing, and a comprehensive report is yet to be issued, but the report highlights the need for greater transparency in the true accuracy of AI generated information.
The results of the final report will serve as a guide for the EU’s member states data protection authorities as they attempt to navigate the new frontier of GenAI.