Artificial intelligence (AI) companion bots are a public health problem. This policy brief proposes applying a public health framework to AI companions. This is a logical extension of a regulatory tradition that has long held companies accountable when their products harm human health. The health risks of AI companions go beyond mental health. They impose physical dangers and make users, particularly young users, at risk for loneliness and isolation. Loneliness has a well-documented impact on physical and mental health. 

While legislative solutions exist, most focus on the first layer of harm: lack of guardrails. AI companions are not designed for medical purposes, but they demonstrably harm public health and should be regulated accordingly. Once bans are recognised as recalls, the real demand becomes clear: companies must prove their products are safe before putting them back on the market.

Key findings

• AI companion bots impose three layers of harm: lack of guardrails, addictive design and risk of disruption of social skills.
• The most comprehensive legislative solutions that address all three of these harms are currently understood as 'bans', and thus often meet resistance.
• They would be better framed as 'recalls', which are already used across many industries where public health impacts are closely monitored. 
• Employing a recall regime in the AI industry can change the playing field by creating a safety floor – so companies can design their AI bots for safety without losing to their competitors.