Article · AI translation
Thanks to IT Home reader 华南吴彦祖 for the tip!
IT Home reported on October 31 that OpenAI said in a blog post published on October 30 local time that keeping communities safe has always been a top priority. Its routine work includes proactively notifying law enforcement of imminent threats of violence. In the third quarter of 2026 alone, OpenAI referred leads on more than 200 cases to law enforcement.
OpenAI pledged to publish a Community Safety Report regularly, systematically disclosing its progress on preventing violence, other community safety issues, and the effectiveness of its safeguards in practice. The first report will be released in early 2027 and will cover safety data from the second half of 2026. Subsequent reports will be published every 6 months.
IT Home learned that OpenAI has added restrictions on how ChatGPT answers questions about past mass attacks and their perpetrators: even when the relevant information is publicly available online, the system will be tightly controlled and provide only extremely brief information. At the same time, people can still use ChatGPT to learn about such events as a whole, their effects on victims and communities, and how to recognize warning signs to prevent future violence.
OpenAI has further tightened its safety policies, strictly limiting detailed accounts of past mass attacks, including specifications of the weapons involved, how they were obtained, and comparisons of the scale of harm that could amplify the perpetrators’ notoriety. These restrictions apply equally regardless of a user’s age or reason for asking.
In addition, when a conversation indicates that a user is becoming fixated on violence or is developing thoughts of seriously harming others, ChatGPT can automatically end the conversation to immediately interrupt that pattern of thought and behavior. If conversations from the same account are forcibly ended multiple times for this reason, OpenAI may temporarily suspend the account.
OpenAI also announced stricter limits on information about guns and other weapons available to teenagers, even when questions appear harmless on the surface. Even if a user shows no violent intent, the system will broadly restrict practical knowledge about weapons and information on bypassing security systems that could be misused to fuel violence.
OpenAI is currently accelerating development of a feature to notify parents immediately when a teenager’s account is temporarily suspended for repeatedly discussing violent content.