OpenAI Faces Pressure From Canada Over AI Safety
The Canadian government is demanding that OpenAI implement stronger safety measures for its models, threatening to impose legislation if the company does not comply voluntarily. The pressure follows an incident where AI-generated chat history was linked to a fatal event. Government officials have stated that, despite recent discussions, OpenAI has not yet introduced what they consider to be substantial new safety features, increasing scrutiny on model oversight and governance.
- The proposed legislation is the Artificial Intelligence and Data Act (AIDA), introduced as part of Bill C-27. If passed, it would regulate the design, development, and use of AI systems across Canada. As of early 2025, AIDA has not been enacted, leading the Canadian government to introduce a voluntary code of conduct for generative AI systems as an interim measure. - The fatal event involved Jesse Van Rootselaar, who killed eight people in Tumbler Ridge, B.C., before taking their own life. OpenAI had banned Van Rootselaar's ChatGPT account in June 2025, seven months prior to the February 10th killings, for policy violations related to violent scenarios. - OpenAI determined the user's activity did not meet its internal criteria for reporting to law enforcement at the time of the ban, stating it didn't identify a credible or imminent threat. The company did, however, proactively reach out to the Royal Canadian Mounted Police (RCMP) after the shooting occurred. - In response, Canadian officials, including Artificial Intelligence Minister Evan Solomon and Justice Minister Sean Fraser, met with OpenAI representatives. The ministers expressed their "disappointment" that the company did not present any substantial new safety measures during the meeting. - This incident is not isolated; OpenAI and other AI developers face scrutiny over other harmful events, including multiple lawsuits in the U.S. alleging that interactions with ChatGPT contributed to suicides. - Canada's proposed AIDA takes a risk-based approach, similar to the EU's AI Act, focusing on regulating "high-impact" systems to mitigate harm and bias. This contrasts with the United States' more sector-specific and innovation-focused regulatory stance. - Prior to the Tumbler Ridge incident, another notable Canadian case involving AI accountability was when Air Canada was held responsible by a small claims court for its chatbot providing inaccurate information about bereavement fares. - In September 2025, before the incident, OpenAI had announced a new framework focused on teen safety, including developing age-prediction tools and implementing stricter content restrictions for users under 18. More recently, in February 2026, the company introduced new safeguards against prompt injection attacks, including an "Elevated Risk" label and a "Lockdown Mode".
