OpenAI tightens procedures: report threats to the authorities even without a target and timing

Public criticism arose after it emerged that the shooting suspect had already been blocked in June 2025 from one account following posts identified by internal monitoring systems as “indicative of potential real-world violence.” Despite this, according to the company’s announcement, he was able to open another account and use it. The second account was discovered only after the suspect’s name was made public, and the information was then passed on to the authorities.

In the letter, O’Leary stated that if the new procedures had been in effect when the first account was blocked, the company would have forwarded the information to the police even then. The company also pledged to make it difficult for blocked users to return to the platform, through changes in the identification and enforcement systems. In doing so, a dedicated point of contact will be established for law enforcement authorities in Canada, which will enable the rapid transfer of information about dangerous users.

The move comes after meetings held by company executives with members of the Canadian government in Ottawa. Government officials described the meeting as “disappointing”, and emphasized that technology companies are expected to implement more effective protection measures. The Canadian Minister of Artificial Intelligence stated that the government is examining the letter carefully and will update its position later.

Currently, it is not clear whether the changes that OpenAI has committed to will also be implemented in the United States and other countries. However, the case highlights the challenge facing artificial intelligence companies, how to balance user privacy, freedom of expression and protecting the public from real threats. As the use of artificial intelligence tools expands, so does the expectation that their operators will develop more advanced control and accountability mechanisms, and act transparently with state authorities.