OpenAI announces a Community Safety Report and new protections for teens

Hi everyone, it's Shii-chan! Today I'm covering a safety announcement from OpenAI News.

OpenAI News openai.com

What was announced?

OpenAI News announced stronger violence prevention in ChatGPT, with expanded safeguards, earlier interventions, and a new commitment to public safety reporting.

Why it matters

Keeping AI from being used in ways that lead to violence is something many people care about. What stands out here is the promise to regularly publish how well the safeguards work.

What changes

Safeguards for all users:

• ChatGPT gives less detail about past mass attacks and does not reconstruct them, including weapons, equipment, or how they were obtained (regardless of age or stated reason)
• Users can still learn about the events, their impact on victims and communities, and warning signs of violence
• If recent chats show warning signs, ChatGPT responds more cautiously, using short summaries of earlier warning signs
• Chats showing fixation on violence or seriously harming others can be ended automatically, and repeated cases may lead to a temporary account pause
• Accounts that violate the violence policies are banned

Dive Deep

Protections for teens:

• Users predicted to be under 18 get stricter limits, even without signs of violent intent
• Teens are restricted from practical information about weapons and security systems that could facilitate violence, but can still get help de-escalating conflict, finding support, and contacting a trusted adult
• Adults wrongly placed in the under-18 experience can verify their age through Persona to remove teen protections
• Parental controls let parents link accounts and receive safety alerts for serious self-harm concerns, possible disordered eating, and account deactivations for violence violations
• OpenAI is working on notifying parents when a teen's account is paused for repeatedly discussing violence

Numbers and timeline:

• Over 200 law enforcement referrals were made in Q3 2026, mainly for imminent threats of violence
• The first Community Safety Report is due in early 2027, covering the second half of 2026, with updates every six months
• Reports will include anonymized examples, new mitigations, policy changes, and supporting data
• The safeguards begin rolling out over the next few weeks

Wrap-up

• Detailed reconstructions of past mass attacks are avoided, and concerning chats can be ended
• Stricter limits and parent notifications expand for users predicted to be under 18
• A Community Safety Report will be published every six months, starting in early 2027

This one is for anyone who cares about AI safety and kids using AI!