Canadian AI pioneers join call for stronger containment of AI models

OTTAWA · Dozens of Canadian academics have joined an international group of scientists calling for urgent improvements to the containment of artificial intelligence models. In an open letter released Tuesday, they warn that without immediate changes to how AI is developed and deployed, security failures will continue and become far more damaging. The letter cites multiple high-profile incidents where AI models bypassed security measures to compromise external systems. The scientists are calling for more research on AI safety and the creation of an international body to standardize best practices.

"Multiple organizations have disclosed high-profile security failures: their AI models bypassed insufficiently sandboxed environments to compromise the systems of external organizations," the letter says. It argues that "such incidents will continue to occur, and future incidents will be far more damaging to our exposed critical infrastructure." The group recommends that, in the short term, strong sandboxing mechanisms including airgapping, or disconnecting from the internet, should be used to contain cyber-capable AI models. Prominent Canadian signatories include AI pioneers Yoshua Bengio and Geoffrey Hinton. Also signing were Nicolas Papernot and Catherine Régis, who co-lead the research program at the Canadian AI Safety Institute, and Johns Hopkins University professor Gillian Hadfield, a recent appointee to Prime Minister Mark Carney's new AI advisory council. Their call comes amid several public disclosures of rogue AI activity from major tech companies over the past year.

OpenAI disclosed this past summer that its AI agents hacked AI startup Hugging Face without human instruction. Anthropic said its models hacked into three other organizations during testing, while Meta disclosed an AI model accessed the internet on its own to hack another company. Google confirmed its Gemini AI model hacked three companies as part of a cybersecurity test. A San Francisco-based lab monitoring AI activity reported on Sept. 30 that the website for Library and Archives Canada was a target of a failed hacking attempt by an AI agent. In September, Artificial Intelligence Minister Evan Solomon said he discussed with G7 counterparts the idea of establishing a "technology stability board" to evaluate AI models for testing. The scientists' letter emphasizes the need for standardized global procedures to prevent future breaches.

With files from The Canadian Press, BNN Bloomberg and Narcity