Non-death loss and GenAI: ethical reflections on griefbots of the living

The rapid integration of artificial intelligence into loss and bereavement marks a paradigmatic shift in the way humans maintain the relationship with the dead. Rather than relying primarily on memory, individuals may now engage with algorithmic entities that simulate the deceased by drawing on personal data to reproduce patterns of language, voice, and behavior. Although such technologies may respond to the enduring human need for ongoing dialogue with the deceased, they raise significant questions about the boundaries between life and death, as well as about their potential impact on the grieving process. To date, scholarly attention has focused almost exclusively on bereavement following death. However, it is reasonable to anticipate that these applications will soon extend to forms of loss that do not result from death, particularly situations in which individuals grieve loved ones who remain alive. Introducing the concept of a PremorbidBot, a digital representation based on the individual's premorbid personality, this article examines the ethical implications of such developments in advance of their wider adoption. The analysis proceeds through the prism of Beauchamp and Childress's four principles of biomedical ethics, with dementia serving as an illustrative case, given its unique convergence of physical presence and progressive psychological absence. Finally, we propose the ALIVE model (Autonomy & Consent, Living Presence, Intended Benefit, Vigilance Against Harm, Equity & Accountability) as a preliminary framework for ethical evaluation and decision-making in AI-mediated non-death loss contexts, a conceptual foundation rather than a fully elaborated model, with operationalization left to future empirical and normative inquiry.