The Canadian government has been granted early access to the latest AI technology from San Francisco-based Anthropic PBC , which has been flagged as a potential threat to financial stability given its power to amplify the speed and precision of cyber attacks by quickly identifying and exploiting vulnerabilities.

“We can confirm that, through Project Glasswing, the Government of Canada has access to Anthropic’s Mythos capability through the Canadian Centre for Cyber Security,” Innovation Minister Evan Solomon’s office said in statement Tuesday.

“This is an important step to help Canada’s cyber defenders better understand vulnerabilities, test systems responsibly, and strengthen protections for government services, critical infrastructure, data and Canadian institutions.”

In a statement, a spokesperson for the CCCS also confirmed the announcement, adding that the agency is also working with OpenAI and its 5.5-Cyber model.

“The Cyber Centre works closely with government partners, critical infrastructure, and the private sector to monitor emerging technology and cyber threats and share guidance that helps strengthen Canada’s overall cyber resilience,” the agency said in the statement. “This includes working with artificial intelligence vendors to better understand the opportunities and risk posed by frontier AI, and their implications for cyber security.”

Previously, only large U.S. tech firms and big banks in the United States such as JPMorgan Chase & Co. had advance access to the AI technology through Project Glasswing, a tightly controlled group of 50 companies and organizations collaborating with Anthropic.

“As AI tools become more powerful, Canada needs to stay ahead of emerging risks,” Solomon’s office said in the statement. “That means working with trusted partners, using advanced tools defensively, and making sure Canadians can have confidence that the systems they rely on are secure.”

On Tuesday, Anthropic announced that Project Glasswing is being expanded to around 200 companies and organizations in total that can use Anthropic’s Claude Mythos Preview artificial intelligence to secure software. Each member must adhere to strict security requirements.

The company said the expansion brings in collaborators in more than 15 countries, most of which provide critical infrastructure. The group now includes industries that weren’t well represented in the initial cohort including power, water, healthcare, communications and hardware, the company said. Anthropic’s collaboration with outside organizations through Project Glasswing includes the U.S. government.

“Many of the new partners are vendors — companies or nonprofits that maintain codebases that are relied upon by lots of other organizations around the world, including governments,” Anthropic said.

“What each partner has in common is that a successful attack on their codebase could be catastrophic. For most partners, we estimate that a major attack could affect more than 100 million people, with important ramifications for both global and national security.”

A spokesperson for the Canadian Bankers Association said the industry group was unaware of which agencies, companies or organizations in Canada are included in the newly expanded Glasswing group. Regulators at the Ontario Securities Commission referred a query to Anthropic and the Office of the Superintendent of Financial Institutions did not respond to a request for comment. The Bank of Canada did not comment directly on an inquiry about Canadian membership in the Glasswing collaboration.

In April, Bank of Canada governor Tiff Macklem said much of the discussion in Canada about Anthropic’s Mythos was taking place in The Canadian Financial Sector Resiliency Group, a public-private partnerships led by the central bank that is designed to strengthen the financial sector’s critical infrastructure against risks.

One member of that group is the CCCS, through the Communications Security Establishment (CSE). Other members of the financial sector resiliency group include OSFI, the Finance Department, TMX Group Ltd. and technology experts from Canada’s largest banks.

Anthropic delayed the public release of its latest AI iteration in April due to its ability to rapidly exploit software vulnerabilities. This kicked off a series of high-profile meeting involving governments, regulators and financial players.

  • Anthropic’s Mythos signals AI risks are coming fast, says Canada’s top banking regulator
  • Canada’s biggest market watchdog warns AI risks like Anthropic’s Mythos need bigger response than traditional regulation

While in Washington that month, Macklem said he had been in touch with United States Federal Reserve chairman Jerome Powell to discuss the risks posed by Anthropic’s latest AI model, with their discussions among a series of high-level communications between Canadian and U.S. officials on the topic.

At the time, he said the Canadian Financial Sector Resiliency Group had met twice by mid-April.