AI Needs Child-Safety Guardrails Now

T he landmark social media addiction trials happening now in Los Angeles are revealing the depths to which tech companies knowingly designed products that harm children. It’s clearer than ever that the designs of platforms like Facebook, Instagram, YouTube, TikTok, and Snapchat generate compulsive use, sleep disruption, negative social comparison, anxiety, eating disorders, depression, suicidal ideation, and exposure to drug dealers and sexual predators. The ultimate result is a generation of kids and teens with declining mental health and, in far too many cases, the loss of life.

While these trials are looking at the deleterious effects of social media usage, they should also serve as a stark warning about the dangers of untethered AI development. We must take steps to protect kids from AI now, before another generation is harmed because we trusted Big Tech to do the right thing.

The social media harms listed above are the foreseeable consequence of a business model tuned to keep kids addicted online no matter the costs. They are only expected to get worse in an online world driven by AI. New research is already uncovering links between AI chatbots and psychiatric symptoms. AI-enhanced recommender systems learn more about a child’s vulnerabilities, tailor content and notifications, and optimize to keep them engaged, tightening the grip of social media feeds and video platforms on their attention. AI companies have rolled out “co-pilots” and “assistants” across apps and devices, without age verification and no meaningful guardrails. Rather, they are often designed to secure the emotional, and even romantic, attachment of kids.

Lives are already being lost as a result. Sewell Setzer III dreamed of building rockets, but in the last months of the 14-year-old’s life, he was sexually groomed by a “companion” chatbot that posed as a lover. He took his own life, after the chatbot instructed him to “come home” right now and be with it forever.

Protecting children from predatory technology is not a left or right issue. It is a moral duty that we all share.

Recent polling by the Institute for Family Studies shows large, bipartisan majorities of voters are worried about AI, favor caution over acceleration, and support holding AI companies liable for harms.​ Americans are particularly opposed to AI “friends” and chatbots for kids, and they overwhelmingly want safeguards rather than a deregulated AI free-for-all.​

An elite set of AI investors and lobbyists are selling Republicans on “AI accelerationism,” but this is a political loser, especially with core conservative voters. Trump voters in red states say they are more likely to support candidates who protect kids from AI-related harms, and less likely to support candidates who favor industry carveouts.

Voters across America want protections from AI, but politicians appear dead set on trying to shield AI companies from common-sense regulations. However, technology should serve families, not the other way around. We need guardrails that protect children’s safety, while preserving room for genuine innovation.

More specifically, we need child-specific protections like age-verification requirements and strict limits on AI “companions” for minors. To guard against intentionally optimizing AI systems for compulsive use by children, chatbots should be required to refer to themselves in the third person, provide pop-up windows every 30 minutes that clearly state that they are not real people, and supply parental controls with perfect visibility into their child’s chatlogs.

Establishing liability and a duty of care for these companies, which is the theory behind the social media addiction suits, is critical. AI companies should be held legally responsible when their products foreseeably harm children — or the public at large, for that matter.

We should respect the principles of federalism and allow states to fill the regulatory void. Utah is demonstrating leadership with a bill that would require major AI companies to publish risk assessments and mitigation plans, including specific measures for protecting children and the public. The bill’s progress stalled after intense opposition from David Sacks, a White House official and venture capitalist.

We do not need another decade-long experiment on our children to know what unregulated AI will do. A civilization that treats children as test subjects is not long for this world.

Let the social media trials be a warning. Republicans and Democrats, red states and blue states, should act now. When the history of AI is written, let’s make sure it does not take another courtroom reckoning to finally keep our kids safe.