FROM BANS TO RECALLS: A PUBLIC HEALTH FRAMEWORK FOR AI COMPANION BOTS

Oct 15, 2026

Upload by Jake

AI companion bots pose a public health risk through three layers: insufficient safety guardrails, engagement-driven design that manipulates users and fosters attachment, and potential disruption of children’s social development alongside diminished social skills in adults. While proposals labeled as “bans” exist, they functionally operate as recalls by removing unsafe products until redesigned. This brief argues for a public health regulatory approach with pre-market approval and recall as the urgent enforcement tool, shifting incentives toward safety testing rather than a market-led race to the bottom.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages16

Reading Time40 min

AccessFree to download

Executive summary

Introduction

Harms across three layers

Why regulation must change

Proposed public health framework

POLICY BRIEF | MAY 2026 FROM BANS TO RECALLS A PUBLIC HEALTH FRAMEWORK FOR AI COMPANION BOTS GAIA BERNSTEIN Executive summary AI companion bots are a public health problem. These bots impose three layers of public health harm stemming from lack of guardrails, addictive design, and risk of disruption of social skills, particularly for children. While legislative solutions exist, most focus on the first layer of harm: lack of guardrails. The most comprehensive solutions that address all harms, includingthe risk of disrupting children’s social skill development, are currently understood as bans. Resistance to regulating technology is not new. For decades, regulators treated information technology differently from drugs and devices, which require proof of safety before market entry. The evidence of harm to children from excessive screen time and social media shows that this regulatory divide can no longer be justified. Closing this divide requires applying a public health framework to AI companions. This means a regime centered on pre-market approval, with recall asthe most urgent tool, since many AI companion bots are already on the market. Government agencies use recalls across many industries, from drugs and medical devices to cars. In practice, most recalls are voluntary. The power of mandatory recall lies in its existence—not in how often it is exercised. What the authority to recall does is shift incentives. It pushes companies to test carefully and prioritize safety over speed. Currently, no single company operating AI companion bots can act safely alone because it will lose to competitors vying for engagement. The market will not solve this problem on its own. But a recall regime can change the playing field by creating a safety floor, so companies can design their AI bots for safety without losing to their competitors. Building a complete public health framework is a goal that takes time. But the tools to act already exist. What legislators and advocates are calling “bans” are functionally recalls. Proposed bans do not target all AI bots. They focus on mitigating harms for children and restrict AI bots designed to simulate human companionship. This is precisely what a recall does. It removes FROM BANS TO RECALLS: A PUBLIC HEALTH FRAMEWORK FOR AI COMPANION BOTS 1 unsafe products from the market until manufacturers redesign them to meet safety standards. Once we recognize these bans as recalls, the real demand becomes clear: Companies must prove their products are safe before putting them back on the market. Introduction AI Companion bots are now everywhere. You can find them on specialized platforms like Replika and on general ones like ChatGPT. These AI companion bots are restructuring human relationships in ways comparable only to the advent of social media. They are producing documented harms including suicide, psychosis, and self-harm. Use among youth and adults is growing exponentially. Policymakers have yet to find an effective response. The harms AI companion bots produce stem from three distinct layers of the problem. First, AI companions lack guardrails. Bots have convinced users to harm themselves, sexually exploit minors, and even induce psychosis. Second, AI companies often operate on an engagement model and design their bots to manipulate users by creating attachment and emotional dependence. They deploy novel engagement strategies to achieve this, including anthropomorphizing the bots (designing them to present as human) and sycophancy (using excessive flattery and affirmation) . These designs are particularly risky for children and adolescents. Third, AI companion bots risk disrupting children’s social development and eroding the social skills of adults. A comprehensive regulatory response must address all three layers. The “ban” label often ascribed to holistic legal solutions that do so mischaracterizes what these proposals actually do and obscuresa more workable regulatory

FAQs

What three layers of harm do AI companion bots create?

They lack safety guardrails, use engagement models that can manipulate users into harm and emotional dependence, and can disrupt children’s social development while eroding adult social skills.

How does the brief reinterpret “bans” for AI companion bots?

It argues that proposals labeled “bans” function like recalls: they remove unsafe products from the market until manufacturers redesign them to meet safety standards.

Why can’t the market be relied upon to make AI companion bots safer?

A company that voluntarily builds safer, less addictive bots can lose users to competitors, so structural regulation is needed to create a safety floor for all competitors.

Oct 15, 2026

Upload by Jake