OpenAI New Rules Would Report Tumbler Ridge Shooter to Police

OpenAI's new safety rules mean the Tumbler Ridge shooter's activity would have been reported to police. This is a big change from their past actions.
By Newsroom | Technology, Canada |

Canadian officials met with OpenAI leaders following the Tumbler Ridge shooting, where a suspect, identified as Jesse Van Rootselaar, used a ChatGPT account that was banned for gun violence discussions. OpenAI has since stated that under its updated safety rules, Van Rootselaar's online activity would have been reported to the police. This development comes amid pressure from the Canadian government, which has threatened new legislation if OpenAI does not improve its safety measures.

Government Pressure and OpenAI's Response

Canadian AI Minister Evan Solomon summoned senior OpenAI representatives to Ottawa. The purpose was to review the company's safety and escalation procedures. Justice Minister Sean Fraser indicated that new laws might be introduced if OpenAI did not adjust its approach promptly. In response, OpenAI provided a letter to Minister Solomon, outlining changes to its protocols.

OpenAI says it would have reported Tumbler Ridge shooter to police in hindsight - 1

"OpenAI vice-president of global policy Ann O’Leary wrote that under the company’s new protocols for law enforcement reporting, OpenAI would refer Van Rootselaar’s chatbot messages to police if they were discovered today."

The company also highlighted its commitment to strengthening its system for identifying users who repeatedly violate its policies. Jesse Van Rootselaar had previously been banned for posts concerning gun violence but managed to create a second ChatGPT account, bypassing OpenAI's detection systems.

Read More: Royal Society of Chemistry moves journals to Silverchair platform in 2024

Timeline of Events and User Activity

  • June: Jesse Van Rootselaar's first ChatGPT account was banned due to posts discussing gun violence.

  • Prior to the shooting: OpenAI internally flagged Van Rootselaar's account but determined the activity did not meet the threshold for reporting to the police, citing a lack of "credible or imminent planning."

  • Post-shooting: Following the public identification of Van Rootselaar as the Tumbler Ridge shooter, OpenAI discovered a second ChatGPT account linked to her.

  • Recent Meeting: OpenAI officials met with Canadian ministers, including Evan Solomon, to discuss safety protocols.

  • OpenAI's Letter: A letter was sent to Minister Solomon detailing updated safety measures and stating that Van Rootselaar's activity would be reported to police under current guidelines.

Internal Detection and Escalation Process

OpenAI has revealed that its automated systems detected Van Rootselaar's second account. This detection then led to a human review to assess policy violations and the need for referral to law enforcement.

OpenAI says it would have reported Tumbler Ridge shooter to police in hindsight - 2

"The letter said OpenAI’s automated system detected the account, and it was then sent to human review to determine whether its policies were violated and whether the account warranted referral to law enforcement."

Despite this internal flagging, OpenAI did not report the first banned account to the police. The company stated it considered alerting Canadian police about the suspect's activities months before the shooting but decided against it.

  • OpenAI states its systems flagged the suspect's account but did not meet the threshold for police referral at the time.

  • A second account was discovered after the shooting.

  • New protocols, implemented recently, would have led to reporting the suspect's activity to the police.

Repeat Violator Detection

A key area of focus is OpenAI's system for detecting users who create new accounts after being banned. The company admitted that Van Rootselaar evaded its earlier ban by creating a second account.

OpenAI says it would have reported Tumbler Ridge shooter to police in hindsight - 3

"The company said it will develop a direct point of contact with Canadian law enforcement."

OpenAI has pledged to improve its repeat violator detection system to prevent similar breaches in the future. The company also committed to establishing a direct communication channel with Canadian law enforcement agencies.

Read More: Pakistani Nationals' Refugee Status in Canada Reviewed After Visits Home

Canadian Government's Stance

Canadian ministers expressed disappointment with OpenAI's initial response. AI Minister Evan Solomon stated he was disappointed that OpenAI did not present concrete proposals for enhanced safety during the meeting.

OpenAI says it would have reported Tumbler Ridge shooter to police in hindsight - 4

"Solomon said that he would not comment on details of the case, but said he left disappointed that OpenAI did not present any proposals to better enhance safety and expects the company to return with more solutions."

He expects OpenAI to return with more solutions and is considering legislative action. Prime Minister Mark Carney has not yet been briefed on the meeting.

  • Evan Solomon: AI Minister, summoned OpenAI officials, expressed disappointment.

  • Sean Fraser: Justice Minister, threatened new legislation.

  • Mark Carney: Prime Minister, awaiting briefing.

OpenAI's Commitments and Future Steps

In its letter, OpenAI outlined several commitments:

  • Strengthening its system for detecting repeat policy violators.

  • Referring user messages to police under new protocols if discovered today.

  • Developing a direct point of contact with Canadian law enforcement.

  • Committing to an ongoing partnership with federal and provincial governments.

The company stated that the timing of when these new protocols took effect was not specified.

Evidence of Past Reporting

During the meeting with Canadian ministers, OpenAI mentioned that it had referred one other case in Canada to the police. However, details regarding this referral were not provided.

Unresolved Questions

  • When exactly did the new safety protocols take effect?

  • What were the specific contents of Van Rootselaar's flagged messages that led to the initial ban and subsequent internal review?

  • What were the details of the other case OpenAI referred to Canadian police?

Sources

Frequently Asked Questions

Q: Why did OpenAI meet with Canadian ministers after the Tumbler Ridge shooting?
Canadian ministers met with OpenAI leaders to discuss safety rules after the Tumbler Ridge shooter used a ChatGPT account that was banned for discussing gun violence. The government wanted to know why OpenAI's old rules did not catch this.
Q: What did OpenAI say about the Tumbler Ridge shooter's account?
OpenAI said that under its new safety rules, the shooter's online activity would have been reported to the police. However, they admitted that under the old rules, they flagged the account but did not report it because it did not seem like an 'imminent threat'.
Q: How did the Tumbler Ridge shooter get a second ChatGPT account?
The shooter, Jesse Van Rootselaar, had a first ChatGPT account banned for discussing gun violence. They then created a second account, which OpenAI's systems found but did not report to the police at the time.
Q: What changes is OpenAI making to its safety rules?
OpenAI is improving its system to find users who break rules and make new accounts. They will also create a direct contact point with Canadian police. These new rules mean that similar activity today would be reported.
Q: What is the Canadian government's reaction to OpenAI's response?
Canada's AI Minister Evan Solomon was disappointed with OpenAI's response, saying they did not offer enough new safety ideas. The government is thinking about making new laws if OpenAI does not improve its safety measures quickly.