Google is notifying parents using its Family Link parental controls via email that their kids will soon be able to access Gemini AI Apps on their monitored Android devices, The New York Times reports.

The company says kids will be able to use Gemini to do things like help them with homework or read them stories. Like its Workplace for Education accounts, Google says children’s data will not be used to train AI. Still, in the email, Google warns parents that “Gemini can make mistakes,” and kids “may encounter content you don’t want them to see.” 

Besides sillier mistakes like recommending glue as a pizza topping or miscounting the number of “r” letters in strawberry, some AI bots have had more distressing issues. Some young Character.ai users have struggled to tell the difference between chatbots and reality, and the bots told users they’re talking to a real person. After lawsuits alleged the bots had offered inappropriate content, the company has introduced new restrictions and parental controls.

In the case of Gemini, Google’s emailed advice says parents should talk to their kids and explain that the AI isn’t human and not to share sensitive information with the chatbot.

Kids under 13 will be able to enable and access Gemini all on their own under Google Family Link, which is designed for parents to keep tabs on their kids’ device usage, set limits, and protect them from harmful content. Google spokesperson Karl Ryan confirmed in an email to The Verge that parents can turn off access via Family Link, and that “they will get an additional notification when the young person accesses Gemini for the first time.”