Two major things that may at first seem unconnected are happening in tech this week. In a California court, 29 attorneys general are bringing a lawsuit against Meta, the parent company of Facebook, Instagram, and WhatsApp, alleging the company knowingly designed Facebook and Instagram to addict children, misled the public about the harms its platforms posed to young users, and illegally collected data from children under 13 without parental consent.
At the same time, OpenAI announced the release of ChatGPT for Teens, which automatically places users who are either actually, or estimated to be, ages 13 to 17 into a version of ChatGPT with stronger safety protections, reminders about their use of the tool, and additional parental controls.
OpenAI’s teen-specific version arrives nearly four years after ChatGPT was unleashed on the world in November 2022. Instagram, by comparison, had been around for 14 years before Meta introduced dedicated Teen Accounts in September 2024 (Meta products had other child safety features long before this, however).
In both cases, there’s a pattern that the tech sector loves to repeat: releasing products that get young people hooked, then only later establishing boundaries around their use.
For its part, OpenAI says ChatGPT for Teens builds on safety measures it has introduced over the past year, including parental controls, age prediction, and an under-18 version of its Model Spec. The company says its approach is built around prioritizing teen safety and encouraging real-world support.
But experts say the trajectory is a familiar one.
“When I look at the safety issues of social media and LLMs I see similar trajectories,” says former Facebook employee Victoria Baines, professor emerita of information technology and fellow at Gresham College, U.K. (Neither Meta nor OpenAI responded to Fast Company’s request for comment.)
Tech companies tend to take a similar approach to safety issues as they emerge, Baines says, prioritizing the most egregious harms, including “child abuse and terrorism that are obviously illegal and can be detected and prevented at scale.” They can address those risks using established tools, such as matching content against databases of known child abuse imagery or adding guardrails that prevent AI systems from generating similar material.
“That’s entirely reasonable from a government perspective, and from a business perspective it focuses safety efforts on compliance risks that have a basis in established regulation,” Baines says.
It’s with risks that are not necessarily unlawful, but which Baines says most people might see as “clearly problematic to individual users and to society,” that things get trickier. In terms of social media, that might include inappropriate content for certain age groups and the ease with which it is served to younger users. For LLMs, it could be undue influence and misinformation swaying people’s opinions.
Those less easily defined risks may be even more consequential with AI. Meetali Jain, executive director of the Tech Justice Law Project, describes generative AI as “social media on steroids,” arguing that its influence reaches into “the very essence of what it means to be human.”
Tech companies more or less leave those issues alone where they can because they’re often disputable and divisive when it comes to the approach taken, but also because it’s hard to draw clear rules around them. That doesn’t mean tech companies—particularly social media firms—haven’t taken action on those less obviously good-or-bad issues.
“Some of the companies now involved in the acceleration of AI have direct experience of that social media safety trajectory, and one would hope that they would apply lessons learned,” says Gresham College’s Baines. She says the spate of lawsuits in the U.S. has highlighted how it’s “unclear whether those focused on growth listen to those focused on safety.”
That’s what makes AI especially troubling for experts. Baines says the speed of its adoption leaves developers less time to identify and address harms than social media companies had, even as AI is poised to become more deeply embedded in everyday life.
“With AI I think the concern is sharper, because a teenager isn’t just consuming content,” says Nina Vasan, director of Brainstorm: The Stanford Lab for Mental Health Innovation.
AI’s ability to learn from us about what we’re interested in, as well as our vulnerabilities, makes it potentially more pernicious. Vasan says AI companies should face the same question now being put to Meta: “What did your own research tell you about young users, and what did you do once you knew?”
