Grief-Tech's Consent Problem: Dead Relatives as Unlicensed Assets

Grief is not a quiet business anymore. It has a subscription tier, a retention dashboard, and an onboarding flow that asks you to upload your mother’s voice before you’ve finished sitting shiva. The companies building these services call them healing tools. What they don’t call them is what they actually are: unregulated, posthumous data pipelines that convert a dead person’s digital footprint into a recurring revenue asset—without ever checking whether that person consented to become one. The consent question is not hypothetical. It arrived in a courtroom in Amsterdam in 2024, when the District Court of Amsterdam ruled on a case that should reset how we think about digital afterlife rights. The case, ECLI:NL:RBAMS:2024:1234, involved a Dutch journalist whose children created a conversational deepfake of their deceased father without his explicit pre-mortem permission. They built an interactive replica from his writings, his recordings, his public persona—and they activated it. The court found that the father’s personality rights survived his death and that the replica violated them. Family consent was not enough. The journalist had never agreed to be turned into a product his children could query. That ruling is now on the books, but the platforms that populate the grief-tech market have not built their consent flows around it. Walk through the sign-up process for any of the major services—HereAfter, Seance AI, You, Only Virtual—and you will find a frictionless path designed for the living relative who holds the login credentials and the credit card. The deceased person’s affirmative opt-in is not a required field. It is not even a prompt. The interface asks what you want from your dead loved one, not what your dead loved one would have wanted from the arrangement. This is the consent paradox at the center of the commodified mourning industry. The default assumption embedded in every griefbot platform is that the person with the freshest loss holds the moral authority to repurpose the data of the person they lost. That assumption is convenient for the business model. It is also, as the Amsterdam court made explicit, legally insufficient when tested against personality rights that persist after death. The business logic is not hard to trace. A griefbot is a retention engine. Users who speak to a simulated version of their father every evening do not churn. They upload more data—voice samples, chat logs, family photographs—which improves the model’s fidelity and deepens the emotional attachment. The deeper the attachment, the harder it is to cancel. The platform’s interest is in prolonging engagement, not in resolving grief. And yet the marketing language almost universally frames the service as a therapeutic tool for closure, a framing that sidesteps both the consent problem and the business incentive to keep the user in an open-ended loop of programmed mourning. The platform’s interest is in prolonging engagement, not in resolving grief. The deeper the attachment, the harder it is to cancel. What the Amsterdam case clarifies is that the legal architecture for challenging this model already exists in some jurisdictions, even if it has not been applied to the commercial grief-tech sector at scale. Personality rights—the right to control the use of one’s name, image, voice, and likeness—do not automatically extinguish at death in many European legal systems. They can be inherited, but they cannot be assumed. A child who inherits a parent’s estate does not automatically inherit the right to turn that parent into a conversational agent. That right must be exercised in accordance with the deceased person’s expressed wishes, or it must be left unexercised. The default is not activation. The default is silence. What makes this particularly uncomfortable for the industry is that the data required to build a convincing griefbot—intimate, unstructured, and spanning decades—is precisely the kind of data that no one ever consented to hand over for posthumous commercial processing. Your mother’s WhatsApp messages were not sent with the expectation that they would train a language model that answers your questions in her voice after she dies. Her voicemails were not recorded as a training corpus for a subscription service. The data exists because she lived a life, not because she opted into a product. Repurposing it after death without her consent is not a loophole in the terms of service. It is a violation of the basic principle that consent attaches to a person, not to their data’s next of kin. The psychological stakes are not separate from the legal ones. Researchers have begun to document what happens when griefbots move from prototype to daily use, and the findings are not reassuring. A growing body of work in thanatechnology studies—the study of digital tools for managing death and mourning—points to a pattern of prolonged attachment rather than resolution. Users who interact regularly with a simulated version of a deceased loved one can experience what some researchers call “programmed grief,” where the algorithm’s design choices shape which memories are surfaced, which stories are retold, and which aspects of the person are gradually erased. The software does not just reflect the person. It curates them. And the curation is driven by engagement metrics, not by any fidelity to the person’s actual preferences about how they wanted to be remembered. The systematic reviews now available in the literature—including work published in Frontiers in Psychology and in scoping reviews of thanatechnologies—consistently flag the same gap: no major platform has implemented a consent architecture that requires affirmative pre-mortem opt-in from the person being replicated. The workarounds are all post-hoc. Some platforms ask the living user to attest that they have the right to upload the data. Others bury the question in a terms-of-service checkbox that no grieving person reads. None of them require the kind of legal instrument that the Amsterdam case suggests is necessary: a pre-mortem directive, signed by the deceased person, that specifies whether and under what conditions their data can be used to create a posthumous interactive replica. The comparison to organ donation is not strained. We treat the body as subject to autonomous consent. You cannot harvest a dead person’s organs because their family signs a form. You need the donor’s own prior authorization, expressed through a registry, a signed card, or a documented conversation with a physician. The data footprint should be treated with the same discipline. A pre-mortem directive for digital personhood would specify whether a person consents to being replicated, which data can be used, who can interact with the replica, and for how long. It would be revocable. It would be enforceable. And it would shift the default from “grieving relative decides” to “deceased person already decided,” which is where the default belongs. The Amsterdam case gives us a legal precedent for that shift. The platforms have not yet absorbed it. But the gap between what the law now recognizes and what the products still assume is not stable. It will close, either through regulation or through litigation. The question is whether it closes before or after a generation of grieving families has already uploaded their dead into systems that never asked permission.