Canada sticks a privacy probe into OpenAI's ChatGPT

Chatbot's overseers accused of collecting, using, disclosing personal info without consent

The Office of the Privacy Commissioner of Canada is investigating OpenAI's generative language app ChatGPT after the watchdog received a complaint claiming the software was collecting, using, and disclosing personal information without consent.

"AI technology and its effects on privacy is a priority for my Office," the country's privacy commissioner Philippe Dufresne declared in a statement this week. "We need to keep up with – and stay ahead of – fast-moving technological advances, and that is one of my key focus areas as commissioner."

Launched last November, ChatGPT went viral as hundreds of millions of netizens flocked to the free tool to generate all types of text. While it may be fun to get the engine to write bad jokes or essay drafts, authorities are growing increasingly concerned about the privacy risks the technology poses.

Italy was the first country to announce it was investigating ChatGPT to see whether the software unlawfully collects data on its citizens, and if the technology could harm minors under the age of 13. The country's Guarantor for the Protection of Personal Data even went as far as to temporarily ban ChatGPT by blocking access to the free online demo. 

Meanwhile, other countries including Germany, France, and Ireland are reportedly monitoring the situation to decide whether ChatGPT violates GDPR rules.

Powered by the GPT-3.5 large language model, ChatGPT was trained on huge swaths of text scraped from the internet. It's likely to have ingested at least some personal information, and officials are concerned that data could be extracted by those querying the bot. OpenAI has already built a filter to stop its older GPT-3 model from spitting out people's phone numbers, for example.

Even though the personal information would probably have to have been publicly posted on the internet for the bot to ingest during training, ChatGPT could potentially be used as a tool to help people find such information more easily. There are other bigger privacy issues too – people might be handing over private, sensitive information to the model when they use it, and it's not clear how OpenAI is storing or using the data.

Some companies like Amazon have warned employees not to share confidential information with ChatGPT – like pasting proprietary code. Concerns that the model could leak data increased when an open source library bug temporarily caused the software to reveal snippets of other people's conversations and payment data.

"The investigation into OpenAI, the operator of ChatGPT, was launched in response to a complaint alleging the collection, use and disclosure of personal information without consent," Canada's Office of the Privacy Commissioner said. "As this is an active investigation, no additional details are available at this time." ®