OpenAI said it considered contacting Canadian police months before a user later carried out a deadly school attack.
The company flagged the account of Jesse Van Rootselaar in June for activity linked to potential violence.
It reviewed whether to notify the Royal Canadian Mounted Police but decided the threshold was not met.
OpenAI said referrals require evidence of a credible and imminent risk of serious harm.
At the time, investigators found no sign of a specific plan.
The 18-year-old killed eight people in Tumbler Ridge, British Columbia, before dying from a self-inflicted gunshot wound.
Victims included a teaching assistant and several students aged 12 to 13.
Police said the attacker had previous mental-health-related contact with authorities.
After the shooting, OpenAI shared information about the account with the RCMP.
The motive for the attack remains unclear.
The case has renewed debate about when technology companies should alert law enforcement.
It also highlights the difficulty of judging risk from online activity without clear evidence of an imminent threat.
