OpenAI has added parental controls for ChatGPT’s Web interface, with mobile controls coming soon. The controls give parents the ability to reduce or remove certain content and dial down personalization by turning off ChatGPT’s transcript memories. At the same time, OpenAI has added the ability to restrict image generation with the launch of Sora parental controls for ChatGPT-connected teen accounts. There are also controls for sending and receiving direct messages through the app. OpenAI says the changes aim “to give families tools to support their teens’ use of AI.” To activate control access, parents must have their own accounts and teens will need to opt in.

“Teens can disconnect their accounts at any time,” The Verge reports, “though parents will be notified if that happens.”

Parents will not routinely have the ability to access their teens’ conversations, even with their accounts linked, though OpenAI explains in a news post that exceptions will be made “in rare cases where our system and trained reviewers detect possible signs of serious safety risk.”

When that happens, “parents may be notified — but only with the information needed to support their teen’s safety.”

Last month, the company generally announced the changes were coming. Now OpenAI has shared the specifics.

Parents are now able to reduce “graphic content, viral challenges, sexual, romantic or violent roleplay and extreme beauty ideals,” in a teen’s feed, and those limitations will be on by default for teen accounts once they’re linked to the account of a parent.

Parents can also decline to allow OpenAI to do model training from teen chats, and implement “quiet hours” when teens can’t access ChatGPT. Other safeguards include the ability to turn off voice mode so teens can only use text chat, and an off switch for image generation.

“If our systems detect potential harm, a small team of specially trained people reviews the situation” and in the case of “signs of acute distress” parents or guardians will be alerted “by email, text message and push alert on their phone, unless they have opted out,” OpenAI explains.

“The safety features come in response to numerous incidents of certain ChatGPT models validating users’ delusional thinking instead of redirecting harmful conversations,” TechCrunch writes, adding that OpenAI has been defending against a wrongful death lawsuit that was filed by a California couple “after a teenage boy died by suicide after months of interactions with ChatGPT.”