OTTAWA - The federal government is dodging questions about whether artificial intelligence companies should be paying Canadian news publishers for content their chatbots are openly using.
Last year, the Liberal government passed the Online News Act, which requires some tech companies to negotiate licensing agreements with news publishers to use their content.
The government won't say if the law should also apply to AI services, such as massively popular chatbots.
Models like ChatGPT, which is operated by Microsoft-backed OpenAI, along with Google's Gemini and Meta AI all admit to using Canadian news to either train or provide answers to users' questions.
It will be up to Canada's broadcasting regulator to determine if AI services are reproducing news as defined in the law, said Heritage Minister Pascale St-Onge's office.
"We are closely monitoring developments in artificial intelligence and their implications for the news media sector," the office said in a statement.
Companies had until last week to notify the Canadian Radio-television and Telecommunications Commission whether the law applies to any of their platforms. That list has not yet been made public.
Meta said it remains exempt from the law because its AI system draws from sources across the web that are not restricted by the legislation.
Google is already set to receive an exemption, after it negotiated a deal with the government to set up a media fund, while Microsoft, which has a stake in OpenAI,didn't respond to a request for comment.
Since the Online News Act passed, AI chatbots have become mainstream, providing human-like communication to users for a variety of tasks like answering questions, giving recommendations or summarizing news articles.
ChatGPT uses publicly available information from a variety of news sources, including Canadian news sites, the chatbot said. This includes includes accessing information from websites, articles and other online resources.
Google's Gemini doesn't directly access and process information from Canadian news sites, the chatbot said. However, news articles can be part of its training data, playing a role in its knowledge and abilities.
The Meta AI chatbot said it uses news sites to help answer users' questions, and was trained on a vast amount of data that includes news articles from Canadian sources.
Prime Minister Justin Trudeau was recently asked on a New York Times technology podcast if the law should be expanded to encompass AI — and get such companies to pay for Canadian news.
He evaded the question, saying the onus is on the platforms to act responsibly.
"What I want is not for government to legislate what platforms should do or not do, because that's a recipe for disaster. We all know how slow governments end up working," he said on the Hard Fork podcast.
"But can we put the onus of leadership and responsibility that goes with it increasingly on platforms — around journalism, around protection of free speech, but also protection against hate speech? Can we find those balances?"
In addition to its Online News Act and Online Streaming Act, which put new requirements on streaming giants, Trudeau's government also proposed a law that would allow companies to be regulated and penalized over their approach to online safety. That bill is still being debated in Parliament.
When it comes to the Online News Act specifically, Canadian Heritage has previously said the law would apply to just two companies: Google and Meta.
The law stipulates that a platform is captured under the law if it shares news or repurposes it on its online platform, along with meeting other criteria.
Google is set to receive an exemption, after agreeing to pay Canadian news publishers $100 million annually, indexed to inflation.
Meta blocked news links from its Instagram and Facebook platforms for its Canadians users in order to comply with the law.
"Our company complies with the Online News Act," a Meta spokesperson said in a statement.
"In Canada, Meta AI draws from sources across the web that are not restricted by this legislation to address questions and provides sources for those results from our search engine partners."
This report by ¹ú²úÓÕ»ó¸£Àû was first published June 17, 2024.
