Minister of Artificial Intelligence and Digital Innovation Evan Solomon on Parliament Hill in Ottawa on Tuesday. Mr. Solomon disclosed that the government is partnering with Anthropic to access its powerful new Claude Mythos model.Justin Tang/The Canadian Press
Artificial Intelligence Minister Evan Solomon disclosed on Tuesday that the Canadian government is partnering with AI giant Anthropic to access its powerful new Claude Mythos model, which is skilled at detecting cybersecurity vulnerabilities.
Mr. Solomon told reporters that, through the Canadian Centre for Cybersecurity, the federal government is now part of Project Glasswing, the name Anthropic gave to its Mythos preview campaign. This will allow some Canadian companies access to the AI model to bolster their cybersecurity.
Anthropic chose to delay the wide release of the AI model because its ability to exploit software vulnerabilities creates the potential for abuse.
Opinion: Canada’s AI strategy is set to fail before it even launches
Instead, the San Francisco-based technology company made a preview version available through Project Glasswing to a select group of companies, including banks and digital infrastructure providers, giving them time to identify system weaknesses in preparation for a new breed of AI-enabled cyberattacks.
“I can confirm that the Canadian government is part now of Project Glasswing, which allows companies to have access to Mythos,” Mr. Solomon told reporters, saying the development is “a very important step for Canadians and for our government to make sure that we’re having access and we can protect our institutions and our countries.”
The AI technology could be used by banks, utilities, stock exchanges and other businesses to test for vulnerabilities, enabling companies to patch security flaws before Mythos is released more widely and hackers get access.
Mr. Solomon said Tuesday he would not give details of how the Mythos preview would roll out.
Opinion: Are we approaching the ‘Silent Spring’ of artificial intelligence?
“Part of what Project Glasswing is all about is testing the systems,” he said. “So that’s really important.”
Experts have cautioned about the cybersecurity risks posed by new AI models such as Mythos, saying that they are able to detect and exploit vulnerabilities to a degree that some describe as dangerous. Anthropic has said its early tests showed Mythos found previously unsuspected flaws in systems that have been around for more than a decade.
Anthropic is expected to release the Mythos model to all customers in the coming weeks, Reuters reported last Thursday.
Mr. Solomon is poised to announce his AI strategy this week, which is expected to include major investments in Canadian AI and moves to consider labelling artificially generated content so Canadians can tell the difference between what is real and what is not.
The Globe and Mail previously reported that Canada’s Communications Security Establishment will soon get access to OpenAI’s latest artificial-intelligence model for cybersecurity testing.
Editor’s note: A previous version of this article incorrectly attributed a statement about potential uses of Mythos to Artificial Intelligence Minister Evan Solomon. This version has been corrected.