Lack of artificial intelligence policies puts GN at risk, privacy commissioner says

Government workers may be unknowingly using artificial intelligence in “unethical ways,” says Nunavut’s commissioner of information and privacy.

“We can be certain that employees of the government of Nunavut are using artificial intelligence tools right now, today,” Graham Steele said Friday, speaking during the second day of Nunavut’s oversight and public accounts committee’s review of the privacy commissioner’s annual report.

The risk, Steele said, is if a GN employee uses AI tools in an ad hoc fashion.

“So another risk that I’ll identify is that because this is so new and has developed so fast, a lot of people are just using it without really understanding how it works,” he said of artificial intelligence technology.

Government employees might not realize they are allowing an AI model access to information that should be kept confidential. Steele said the Government of Nunavut is vulnerable because of a lack of policy around how it deals with artificial intelligence both internally and externally.

“There’s a risk that employees of the Government of Nunavut will use it in wrong ways, that they’ll use it in unethical ways,” he said.

Some AI models are seemingly free — in order to provide a service, a model may require access to private information on the user’s computer or retain confidential information that has been provided to it.

Artificial intelligence is not a bad thing, Steele said, and there is an opportunity to leverage artificial intelligence at the GN. However, there needs to be guidance and policies around its use.

“We desperately need some guidelines about how to use it properly and ethically,” he said.

“To give an example, the government of Ontario has issued policy guidance to the entire public service of Ontario about the ethical use of artificial intelligence, when it should be used, when it should not be used, how it should be used.”

The Nunavut government is actively working on developing a policy around the use of artificial intelligence, said Mark Witzaney, director of access to information and protection of privacy.

“There’s been significant work done to look at this from a policy perspective, from a research perspective,” Witzaney said Friday.

He said there is no timeline, but official guidelines will likely be complete later this year or early next year.