Canadian privacy regulators have concluded that OpenAI’s early approach to training ChatGPT did not fully comply with the country’s privacy laws. The nation’s watchdog has also added to growing global scrutiny over how artificial intelligence companies collect and use personal data.
The allegations come from the collaboration between the privacy commissioner in Canada and privacy officers in Quebec, British Columbia, and Alberta. The investigation was carried out after a complaint that OpenAI had collected and analyzed personal information without their consent in building ChatGPT.
The core issue in this case is the volume of personal data used in training large language models. In the course of investigations, it was established that OpenAI uses large sets of data, some of which might contain personal information collected from public sites online.
As per the report, some of that personal data contained sensitive information regarding the individual’s health, political affiliation, or even children.
The regulatory body believed that many of those individuals were unaware of their data collection and analysis, thus making the consent meaningless.
Regulators flag data collection, transparency, and user control concerns
The investigation found a number of issues, including the collection of too much personal information, a lack of transparency as to how the information was used, and limited opportunities for individuals to access, correct or delete information about themselves.
The officials also criticised what they described as weaknesses in the accountability and privacy governance practices of the early development of the chatbot by its creators at Open AI.
Canada’s Privacy Commissioner Philippe Dufresne said the case illustrates the difficulties that arise when rapidly moving technologies outpace current protections.
Regulators said innovation and privacy protection should go hand-in-hand, saying companies building powerful AI systems need to build stronger protections into their products from the outset, not after launch.
The company disagreed with some of the findings but said it has already made a number of changes to improve privacy protections.
It has reduced the amount of personal information used to train, implemented tools to detect and filter personal data, and phased out older models developed under practices regulators found problematic.
Officials acknowledged that OpenAI has taken steps to address some of the concerns raised during the investigation. The latest versions of ChatGPT contain more safety features and provide users with more information about how their data might be used.
Case comes as OpenAI’s global scrutiny rises
The issue comes against the backdrop of worldwide struggles among governments trying to figure out how AI companies should reconcile their quest for innovative ideas with personal privacy.
Regulators have been more vigilant about the data AI technologies rely on as they become a part of everyone’s lives.
The investigation serves as a wake-up call to policymakers that while the era of AI presents a different set of challenges when it comes to privacy issues, the existing rules will have to be observed.
The findings do not bar ChatGPT’s services from being available in Canada, but they demonstrate that AI companies should prepare themselves for a higher degree of regulation in the coming years.
The outcome may also affect future discussions concerning AI regulation in general.
