EU chatGPT releases too much fake information to comply with laws.
The European Data Protection Board (EDPB) has published the first progress report from the “GPT Task Force” and the news is not good for ChatGPT.
OpenAI's efforts to bring its main AI model, ChatGPT, into line with EU laws, including the General Data Protection Regulation (GDPR), were recognized by the EDPB but ultimately deemed insufficient.
According to the EDPA document:
Although the measures taken to comply with the transparency principle are important to avoid misinterpretation of ChatGPT results, they are not sufficient to comply.
The findings come as OpenAI spent most of 2024 dealing with temporary suspension orders from several European member states.
As Cointelegraph reported in January, Italy's data protection agency found ChatGPT and OpenAI still in violation of Italian and EU data privacy laws despite warnings and subsequent bans in March 2023.
According to the EDPB, OpenAI has not done enough to align ChatGipt with EU rules since its launch.
The main complaint seems to be that ChatGPT is prone to extracting inaccurate information.
“In fact, due to the inherent nature of the system, current training approaches lead to models that produce biased or distorted results,” writes EDP.
The report also expressed EDPB's concern that “results provided by ChatGPT may be perceived as accurate by end users.”
It is currently unclear how OpenAI will bring ChatGPT into compliance. A GPT-4 model, for example, contains billions of data points and trillions of parameters. Under GDPR standards, it is impossible for people to combine the database to ensure accuracy at a level that is reasonably accurate.
Unfortunately for OpenAI, the EDPB clearly writes that “specifically, technical incapacity cannot be invoked to justify non-compliance with these requirements.”
Related: OpenAI Bans ChatGPT Voice Accused of Impersonating Scarlett Johansson
