Minister of Artificial Intelligence and Digital Innovation Evan Solomon speaks on Parliament Hill in Ottawa, on Feb. 23.Justin Tang/The Canadian Press
Federal Artificial Intelligence Minister Evan Solomon said that Anthropic’s decision to not publicly release its powerful new AI model is the “responsible” approach, after meeting with officials from the company on Tuesday.
The San Francisco-based company announced its Claude Mythos model earlier this month, but opted not to make it widely available because of the cybersecurity risks it presents. Adept at reasoning and writing computer code, Mythos is also skilled at finding and exploiting vulnerabilities in existing software, according to Anthropic.
The company has instead opened preview access to tech giants such as Amazon, Microsoft, Apple, Google, CrowdStrike, Palo Alto Networks and JPMorganChase to first shore up their defences.
Opinion: Canadian companies need access to Anthropic’s Mythos before hackers arrive
“Anthropic’s approach of working with defenders first, rather than releasing this new model broadly, is the responsible path and gives people protecting critical systems a head start,” Mr. Solomon said in a e-mailed statement. “This is the kind of proactive approach we expect from frontier AI companies.”
Mr. Solomon also cited the Canadian Centre for Cyber Security (part of the Communications Security Establishment) and the Canadian AI Safety Institute, or CAISI, as organizations capable of identifying and responding to the risks posed by AI.
It is not clear if any Canadian companies or organizations have access to Mythos. Mr. Solomon’s office did not immediately respond to follow-up questions on the matter. Anthropic also did not respond to a request for comment.
Even as Mr. Solomon welcomed Anthropic’s approach, some experts in Canada are alarmed that the decision to publicly release powerful AI models rests with commercial entities in the absence of a clear regulatory and auditing process. Canadian AI pioneer Yoshua Bengio previously told The Globe and Mail that dynamic is “deeply concerning.”
Some experts have also called for Anthropic to extend Mythos access to Canadian institutions.
Researchers working with CAISI, for instance, would be well-positioned to assess Mythos. The institute launched in November, 2024, with $50-million in funding over five years from the Trudeau government. CAISI runs a program to study the immediate and long-term risks of advanced AI models through the Canadian Institute for Advanced Research and is housed within Innovation, Science and Economic Development Canada.
Opinion: Mythos sets the world on edge. What comes next may push us beyond
CAISI’s counterpart in Britain, the AI Security Institute, published an assessment of Mythos on Monday and found that it was more capable than other models at autonomously exploiting software vulnerabilities, particularly in weakly defended systems. Researchers could not say for certain whether Mythos would be able to attack more robust systems, however.
Canadian bank executives and regulators met on Friday to discuss the cybersecurity risks posed by Mythos. The meeting of the Canadian Financial Sector Resiliency Group, which is chaired by Alexis Corbett, the chief operating officer of the Bank of Canada, followed similar discussions in the United States last week.
Canada’s Communications Security Establishment did not comment on Mythos specifically, but said that AI models represent a growing threat when used by bad actors.
“These models can accelerate how quickly vulnerabilities are identified and exploited, including flaws in software code and weaknesses in existing security controls,” spokesperson Janny Bender Asselin said in an e-mail.