Griefbots are AI personas produced from a deceased person’s digital data such as their social media content, text messages, photographs, videos and audio recordings. This digital data is used to create a “dynamic digital replica” that personifies a dead person, capable of learning and adapting to prompts and cues.[1] Such replicas range from text-based chatbots to full AI avatars designed to reproduce a person’s speech patterns, humour and personality. It is possible for someone to create such a replica of themselves before they die, while in other circumstances loved ones rely on the data available online about and of the individual and those held in their own records post-their loved one’s death.[2]
The use of griefbots raises concerns about psychological wellbeing, accuracy, privacy, commercialization and the treatment of a person’s identity (and dignity) after death. The principal concerns associated with griefbots can be set out in five overlapping categories, each of which raises distinct questions about their design, use and regulation:
- Psychological wellbeing and prolonging the grieving process. Interaction with a digital replica may offer users a means to express thoughts or feelings about or to a deceased person that remain unresolved.[3] The false sense of companionship offered by a griefbot is inherently risky for people experiencing prolonged grief or those who retreat into isolation as they grieve.[4] A user may become increasingly engaged with the griefbot at the expense of substituting a simulated interaction for human support.[5]
- Accuracy, authenticity and the risk of distortion. A griefbot does not preserve the deceased person’s consciousness or continue that person’s lived development; instead, it produces responses from available data and from the user’s ongoing interactions with the griefbot.[6] The griefbot may hallucinate plausible but inaccurate statements, creating new “memories,” opinions or responses that the deceased person never had or expressed.[7]
- Privacy and consent. Building a convincing griefbot requires extensive personal information. Submitting these materials to a large language model creates risks concerning how the information is collected, stored and reused.[8] Consent is particularly difficult after death, especially if an individual has not included specific instructions in their Will regarding the use of their digital data. An individual may create a replica using another person’s information even if that person never agreed to the creation or use of their digital likeness.[9] This creates a tension between the wishes of surviving loved ones and the deceased person’s autonomy over their identity and personal information.[10]
- Commercialization, control and ownership. Griefbot providers may gain access to uniquely intimate information about the users’ behaviour, relationships and vulnerabilities, creating opportunities to profit. Service terms may also affect who controls the underlying data and the resulting replica.[11] These issues make it important to distinguish between ownership of source materials, control over the generated model and the appropriate safeguards that ensure a person’s personal information is used as they consented to while they were alive.
- Dignity and post-mortem rights. The protection and respectful treatment of deceased persons has been framed as a human-rights concern, but there is not a unifying set of rules or guiding principles across the globe (or even across a handful of countries) for safeguarding the dead.[12] Griefbots create an additional form of post-mortem exposure because a person’s likeness may continue to speak, respond and appear in new contexts after death. Without clear limits, a replica may therefore be used in ways that conflict with the deceased person’s values, wishes or dignity.
The current Canadian privacy framework is fragmented; Canadian legislation provides no clear mechanism directed specifically at preventing a replica from being made without the subject’s consent, and the federal Personal Information Protection and Electronic Documents Act does not expressly establish privacy rights after death.[13]
In the absence of comprehensive statutory rules, fulsome and thoughtful estate planning offers somewhat of a safeguard. Testators can include a clause in their Will respecting the use or non-use of their digital data post-mortem. Such a clause could contain instructions for a person’s digital footprint and broad powers for estate trustees to manage it.[14] When making a Will, the testator should consider how they do or do not want their digital legacy handled and contemplate the use of clear provisions that authorize or prohibit AI uses, including directions concerning name, image, likeness and voice. However, these measures remain incomplete because extensive recordings may be held by third parties and legal recognition of post-mortem publicity rights is inconsistent.[15] As griefbots and other forms of agentic AI continue to develop, the legal landscape will likely place greater emphasis on privacy and data protection of the dead, with clearer standards needed to ensure that the collection and post-mortem use of personal information respects consent, dignity and individual privacy rights.[16]
[1] Inside Halton: Most Canadians don’t want replicas for their dead loved ones, but the law has no clear answer to stop ‘Griefbots’. (link)
[2] UofT – Schwartz Reisman Institute for Technology and Society: From mourning to machine: Griefbots, human dignity, and AI regulation. (link)
[3] CNN (YouTube): Grief bots allow people to talk to the dead. Does it help? (link)
[4] Ibid.
[5] CTV News: AI ‘griefbots’ offer comfort, but are raising concerns. (link)
[6] CNN (YouTube): Grief bots allow people to talk to the dead. Does it help? (link)
[7] CTV News: AI ‘griefbots’ offer comfort, but are raising concerns. (link)
[8] Inside Halton: Most Canadians don’t want replicas for their dead loved ones, but the law has no clear answer to stop ‘Griefbots’. (link)
[9] UofT – Schwartz Reisman Institute for Technology and Society: From mourning to machine: Griefbots, human dignity, and AI regulation. (link)
[10] Ibid.
[11] CNN (YouTube): Grief bots allow people to talk to the dead. Does it help? (link).
[12] United Nations: Protecting and respecting the dead makes us human (link)
[13] Personal Information Protection and Electronic Documents Act, S.C. 2000, c. 5. (link)
[14] Inside Halton: Most Canadians don’t want replicas for their dead loved ones, but the law has no clear answer to stop ‘Griefbots’. (link)
[15] The Estate Registry: Posthumous AI avatars are here. Estate planning needs to adjust.(link)
[16] Inside Halton: Most Canadians don’t want replicas for their dead loved ones, but the law has no clear answer to stop ‘Griefbots’. (link)
The information and comments herein are for the general information of the reader and are not intended as advice or opinion to be relied upon in relation to any particular circumstances. For particular application of the law to specific situations, the reader should seek professional advice.
