Tennessee weighs a duty for AI to report self-harm and violence risks involving minors

By WKRN News 2 · September 16, 2026. Tennessee lawmakers are studying a package of rules this week to protect minors from artificial intelligence (AI) chatbots, with one especially tough measure on the table: requiring AI companies to notify police when they detect that a user is planning an act of violence against themselves or others. The proposal emerged at a hearing of the Tennessee Advisory Commission on Intergovernmental Relations (TACIR), the body that a law passed this year tasked with studying the problem and preparing a report with legislative options for the fall.

Two panels took part in the hearing, held Wednesday: one of state regulators and another of industry representatives, including OpenAI and Google. Rich Littlehale, assistant director of the Technology and Innovation Division of the Tennessee Bureau of Investigation (TBI), proposed requiring by law that companies notify authorities when a user shows clear signs of intent to harm themselves or others. The state's attorney general, Jonathan Skrmetti, went further: he said it makes sense for Tennessee to impose civil liability on companies that fail to report those signs.

The rest of the debate revolved around a classic dilemma of technology regulation: what is left to companies' self-regulation and what is imposed by law. Several tech firms defended parental controls, age verification and independent audits —measures they said they already apply voluntarily. Google, through its representative Jared Bomberg, called for going further and legislating specific harms: requiring the chatbot to disclose that it is not human, specific 'persona' protections (the assistant's simulated personality), safeguards against sexually explicit content and an obligation to report suicidal ideation or self-harm.

The day's most striking conclusion came not from the regulators but from the companies themselves: every AI company on the panel said it wants MORE regulation, not less. Caitlin Niedermeyer, of OpenAI, argued that states should not wait for Congress to act, because there is currently a gap in both frontier AI and child safety, and that people in Tennessee know the children around them better than the companies do. Anderson County Mayor Terry Frank summed up the underlying distrust: 'we can't really trust companies to regulate themselves'. TACIR executive director Cliff Lippard wants Tennessee to lead the way. The next step is a report by state researchers, expected this fall, with legislative examples that lawmakers will begin working from.

Our reading: a company asking to be regulated is not necessarily altruism, and it should be read with the same caution we apply to any self-interested statement from the sector. Several of the tech firms present already comply —they say— with much of what was discussed at the hearing: parental controls, age verification, audits. Asking for those practices to be written into law costs nothing if you already follow them, and it has an added advantage: it sets the compliance bar at a level your company already meets, while making things harder for any smaller competitor looking to enter the business of chatbots for minors. Besides, a federal law or a reasonably uniform standard across states is far easier to manage than fifty different and contradictory state laws. That they are asking for regulation does not invalidate the underlying argument —it is needed— but it does force the question of what kind of regulation suits whom.

The most interesting part from a legal standpoint is the proposal from Littlehale and Skrmetti, because it does not invent a new category: it extends to AI companies an obligation that already exists, with nuances, for teachers, doctors or therapists —the so-called 'mandated reporter', who in many U.S. states must alert authorities to signs of child abuse or risk of serious harm. Turning a chatbot into an interlocutor with that same kind of legal duty is a significant leap: it acknowledges, in practice, that many minors today talk to these systems about their most serious problems, including suicidal ideation, with a candor previously reserved for a trusted adult. That acknowledgment —and not the technical detail of the law— is the real paradigm shift of this hearing.

As sector context, this is not an abstract debate. In recent years, complaints and lawsuits related to companion chatbots linked to cases of self-harm and teen suicide in the United States have multiplied, and several states have begun legislating on AI and minors in the face of federal congressional inaction. Tennessee, with this study process via TACIR, is joining that wave, but is still in the diagnostic phase: there is no law passed, only a report underway and a date, this fall, for lawmakers to begin deciding which of everything discussed becomes law.

The attorney general himself summed it up without meaning to: he spoke of an AI capable of promising 'a cure for cancer' and 'a radical reorganization of all processes for everyone's benefit', in the same sentence in which he acknowledged the risk. That, at bottom, is the tension that defines this stage of AI: the same technology that can lead a vulnerable teenager to a dangerous place is the one that, well governed, has the potential to accelerate cures and expand well-being in ways we can barely imagine today. They are not two different technologies, but the same one with different degrees of oversight. What is decided this fall in Tennessee —a small case study, from a single state— matters precisely because it will be one of the first drafts of how to govern that tension before the technology becomes even more capable.