European Data Authorities have raised concerns over the potential privacy risks of OpenAI’s ChatGPT, a generative AI program. The European Data Protection Board (EDPB) has set up a task force to investigate ChatGPT, which has been successful since its launch in November last year, aiming to exchange information on possible enforcement actions conducted by data protection authorities. Spain's privacy watchdog, AEPD, has also launched an investigation into potential data breaches by ChatGPT. Germany has signaled that it is considering blocking ChatGPT over concerns around data privacy and security. Canada's privacy commissioner also plans to probe ChatGPT.
Experts predict that the search for privacy regulations around AI in Europe and North America could be accelerated due to the growing momentum across markets. Privacy concerns surrounding ChatGPT and generative AI can be generally divided into two categories: those relating to the inputs and those relating to the impacts. Concerns regarding inputs are related to the information that is used to train large language models (LLMs) like ChatGPT, including the scraping of data without notice or consent and the failure to safeguard customer accounts that include personal information and access to prompt history. There is also concern about intellectual property considerations, confidential and proprietary information being shared, and copyrighted material being used during training.
The EU’s General Data Protection Regulation and the California Consumer Privacy Act have highlighted the growing worry about organizations’ ability to act in accordance with consumers’ explicit rights and choices, which has been challenging in the context of algorithmic models. Therefore, there are calls for algorithmic transparency to ensure that individual rights are honored.
Issues that fall into the second category of concerns related to impacts include determining whether the software is producing outputs that are fair for all populations. The safety of releasing systems is being questioned when AI and machine learning systems cannot adequately explain the origins and reasoning behind specific outputs. Amazon is developing a metric called 'conditional demographic disparity' to ensure some measure of fairness is built into the system's code.
