OpenAI has taken British Columbia’s Office of the Information and Privacy Commissioner to B.C. Supreme Court, with the goal of quashing the office’s investigation into aspects of the Tumbler Ridge school shooting.
In a petition filed earlier this week, the artificial intelligence giant asks the court for judicial review of the OIPC’s order demanding that it provide answers and documents to support the investigation.
The petition asks the court to quash the investigation in its entirety or, alternatively, to find the order unreasonable, quash it and set it aside.
The court filing does not reproduce the list of questions and documents the OIPC sought to compel OpenAI to provide, and the office provided few details in a statement acknowledging the investigation this week.
The OIPC said its investigation is “related to the use of personal information by OpenAI in context of the Tumbler Ridge shooting.”
“Due to the sensitive nature of the subject matter of this investigation, the OIPC is limiting information disclosed publicly to what is strictly necessary,” the office’s statement reads.
It goes on to explain that the OIPC “became concerned” after seeing media reports about 18-year-old shooter Jesse Van Rootselaar’s use of OpenAI’s chatbot ChatGPT.
Van Rootselaar is reported to have queried the chatbot about mass shooting scenarios, engaging it in “lengthy, violent fantasies about mass killings.”
After Van Rootselaar’s account was banned for violating OpenAI’s policies, the teen made a new one. ChatGPT then allegedly provided the shooter with advice on how to “never get flagged again” by describing situations as fictional or hypothetical.
In September, the previous B.C. government filed a lawsuit against OpenAI over the company’s failure to notify law enforcement about Van Rootselaar when it banned the first account.
According to the OIPC, its investigation is focused on “whether the use of personal information (including as contained in and derived from prompts, uploaded content, usage data, interactions with other users) in the delivery of the service (i.e. ChatGPT responses) is one that a reasonable person would consider appropriate in the circumstances, including the potential for responses to influence behaviour.”
OpenAI’s judicial review petition describes the OIPC’s order as “sweeping,” “overbroad” and “unnecessary.”
The petition describes the OIPC’s requests as falling into four categories: “background information,” “age assurance,” “response generation and personalization” and “content review and moderation.”
OpenAI argues that the OIPC did not have “reasonable grounds” to open its investigation, and that the company’s use of Van Rootselaar’s personal information has already been reviewed by the OIPC in the form of a joint investigation it conducted with other privacy commissioners from across the country.
The company argues that the OIPC’s investigation is outside the scope of its authority, and notes that the office’s order appears to contradict the provincial government’s claims in its lawsuit.
“The OIPC provided no reasonable grounds to believe OpenAI’s review (of Van Rootselaar’s accounts) constituted a breach of (B.C.’s Public Information and Protection of Privacy Act),” the court filing reads.
“The implication, however, appears to be that OpenAI improperly used Ms. Van Rootselaar’s personal information, thereby breaching Ms. Van Rootselaar’s privacy rights under PIPA, by conducting that review. The province of British Columbia has publicly sued OpenAI, alleging the opposite—that OpenAI should have made greater use of Ms. Van Rootselaar’ s personal information by disclosing her chats to law enforcement.”
