Scientists warn that more damaging hacking incidents targeting AI systems are likely unless stronger safeguards are adopted. An international group of researchers is calling for improvements to how AI models are contained and managed, arguing that current practices leave systems vulnerable to misuse.
Several outlets report that dozens of Canadian academics are among the signatories. The warning focuses on preventing harmful access to or deployment of AI, and it urges changes to containment measures rather than relying solely on reactive security after incidents occur. The accounts emphasize the researchers’ concern about future risk and the need for technical and governance steps to reduce the likelihood that attackers can exploit AI capabilities.
While the coverage varies slightly in how it frames the issue—such as referencing Ottawa when describing the Canadian involvement—the core message is consistent across outlets: researchers want clearer, stronger containment standards for AI models to help limit hacking and harmful outcomes.
