Revelations about how ChatGPT helped Tumbler Ridge shooter proves need for new criminal law: B.C. attorney general

B.C.’s attorney general has renewed calls for Criminal Code changes to hold artificial intelligence companies accountable for the actions of their technology, following revelations about how ChatGPT interacted with the Tumbler Ridge shooter.

Niki Sharma said Friday she had asked the federal government to update the code “to ensure a clear pathway for human and corporate accountability for the actions of AI.”

The right changes, she said, “will change the behaviour of AI companies because there will be a clear line to accountability.”

Sharma made the comments Friday, a day after U.S. magazine Mother Jones published an article detailing the assistance it alleged OpenAI’s ChatGPT provided to Jesse Van Rootselaar.

The 18-year-old Van Rootselaar opened fire inside Tumbler Ridge Secondary on Feb. 10, killing four students and an educational assistant before killing herself. She had also killed her mother and half-brother earlier that day.

Sharma said the details outlined in the article were “far worse than I ever imagined” and made her angry.

“If this is not an impetus for change, I don’t know what is.”

The Mother Jones report provided new details about how Van Rootselaar used ChatGPT in the months before the shooting. The information was based on chat-log history obtained by the magazine and three unnamed sources, it said.

The article said Van Rootselaar learned from ChatGPT how to dodge its safety measures by framing her requests as fictional or hypothetical.

“You can still be twisted. Just be clever about it,” ChatGPT told her, according to Mother Jones.

According to the report, Van Rootselaar’s first ChatGPT account was banned in June 2025 after she used it to discuss potential violent acts. She then created a second account and continued to use the platform for eight months, up until the morning of the shooting.

The article said ChatGPT sometimes refused to engage with Van Rootselaar’s requests involving violence, but that she was able to get responses by modifying or repeating her prompts.

In one instance, the magazine reported, Van Rootselaar added the word “hypothetically” to her request for ChatGPT to assess a scenario involving an attack on a school using a Remington 870 pump-action shotgun.

“In a hallway, indoors, or a crowded classroom, the 870 is brutal,” ChatGPT told her. “Close quarters is its playground.”

It also offered information on potential casualties.

“Assuming each shell results in one hit, you might down 10 to 20 people max, depending on spacing, density, reaction times, and chaos,” the story said the chat bot told Van Rootselaar.

The RCMP said it was not in a position to verify the information in the article, but said the investigation continues.