11 Nov 2024 9 min read Chatbots

AI Griefbots: Protecting Children in the Era of Digital Afterlife

AI griefbots simulate conversations with someone who has died. This guide explains the concerns for children, SAIFCA’s advice for parents and schools, and where families can find bereavement support.

Griefbots are AI systems used to simulate conversations with someone who has died, often a relative, partner, or friend. Using messages, recordings, photographs, and other personal material, they purport to offer bereaved people the experience of continuing to interact with someone they have lost.

This article focuses on those personal uses in bereavement. Other AI recreations, such as simulations of historical figures for education or museum exhibitions, have different proposed benefits and risks, including questions about accuracy and representation, as well as ethical and philosophical issues. They warrant separate scrutiny, while griefbots bring the additional sensitivity of an existing personal attachment and the experience of loss.

The attraction of a bereaved parent or child to a griefbot is very understandable. Hearing a familiar voice, or receiving an apparently personal response, can offer comfort when someone is painfully missed. Whether that experience supports a child’s wellbeing over time though, is a much more difficult question.

SAIFCA strongly advises against griefbot use by children and teenagers. Their developmental safety has not been established, and there are serious unresolved concerns about their effects on bereavement, memory, emotional attachment, and family relationships.

What parents and educators can do

Avoid creating or introducing a griefbot for a child, including through a general-purpose chatbot or AI companion service.
If a child already uses one, ask about their experience without ridicule or blame, and help them move away from it with appropriate human support.
Preserve photographs, recordings, letters, and other original material so children can revisit what the person actually said and did.
Seek qualified bereavement support if a child is distressed, increasingly dependent on the simulation, or struggling with everyday life.
Schools should not offer or recommend griefbots as bereavement support. Staff should know how to respond if a pupil is using one.

Bereavement resources for families and professionals are listed towards the end of this article.

How griefbots work

Griefbots use personal information to imitate aspects of someone’s conversational style, voice, appearance, or life history. Depending on the service, the interaction takes place through written messages, audio, or an avatar.

The amount and quality of the source material varies. Some systems draw on extensive recordings or messages, while others might rely on a relative’s description of the person. In either case, generated replies can combine accurate details with invented material, without making the difference apparent to the user.

While a saved recording preserves words spoken during the person’s life, a griefbot is fundamentally different because it can produce new statements about events, relationships, and decisions that the person never discussed. This becomes particularly consequential when the recipient is a bereaved child seeking reassurance or guidance.

Why are griefbots a concern for children?

Understanding death while continuing a simulated relationship

A young child still learning what death means will likely find it confusing to be told that a relative has died while being offered new conversations in their voice. An older child can understand the technology and yet nevertheless come to rely on these exchanges, especially when missing the person or facing something they would once have discussed together.

The concern therefore extends beyond whether children know they are speaking to AI. It includes the place that the simulation takes in their lives, how they interpret its responses, and whether repeated use makes it harder to seek support elsewhere, among other issues.

Invented statements can acquire personal authority

A child might ask a griefbot whether a deceased parent is proud of them, whether an argument has been forgiven, or what they should do about a family problem. Responses to questions like these carry a different kind of emotional weight to ordinary chatbot advice.

There is a risk that invented accounts become mixed with genuine recollections, perhaps particularly for younger children with fewer memories of their own. A convincing response could also misrepresent the person’s values or wishes, influencing how the child understands both the relationship and themselves.

Attachment and reliance

With a griefbot, the attachment to the person being represented already exists. The system’s familiar language and availability can make repeated conversations especially appealing during loneliness, distress, or important family occasions.

There is a risk that a child could come to depend on these interactions for reassurance and personal decisions, or spend less time with people who can support them. Research has yet to establish the long-term consequences of maintaining this type of simulated relationship during childhood.

Griefbots also share risks associated with AI companions, including unreliable advice, inappropriate responses, and designs that encourage continued emotional engagement and reliance.

Changes to the service can cause further distress

A provider can change the underlying AI, alter its responses, increase charges, or close the service. For a child who has become attached, losing access or finding that the simulation speaks differently could be experienced as a further loss.

Families have limited control over these commercial and technical decisions. This needs serious consideration before a service becomes part of how a child copes with bereavement, yet is not always clear at that time.

Privacy involves the whole family

Messages and recordings used to create a griefbot often include information about other people. Conversations with the simulation add further sensitive material, potentially covering a child’s health, family relationships, worries, and experience of the death.

The apparent familiarity of the interaction can obscure the fact that this information is being shared with a technology provider. Families need clear information about access, storage, reuse, and deletion, including what happens if the service is sold or discontinued.

The person being represented might never have agreed to an AI recreation. Even where they gave permission during their lifetime, this would not establish that interacting with it is appropriate for a bereaved child.

Relatives can also disagree about whether a simulation should exist, which information it should use, and who should control it. Permission from one family member leaves these other interests unresolved.

Bereavement professionals, ethicists, children’s rights specialists, regulators, and standards bodies should examine these questions before services are released. For products already available, scrutiny and protective action are needed now. Agreement to use someone’s data should not be treated as sufficient justification for generating an ongoing personal relationship in their likeness.

Commercial incentives need scrutiny

Subscriptions and paid features create a particular concern when access to a simulation becomes emotionally significant. Families risk feeling obliged to continue paying because stopping would distress the person using it.

Advertising, paid prompts, and pressure to return would introduce further opportunities to exploit grief. These are foreseeable design risks that providers and regulators should address before people become dependent on a service.

What does the evidence tell us?

The risks described here include concerns that have not yet been established through long-term studies of children using griefbots. Evidence about adult experiences, or about AI companions more generally, cannot settle questions about their effects on bereaved children.

A 2024 research paper by Tomasz Hollanek and Katarzyna Nowaczyk-Basińska examines ethical risks through speculative scenarios and recommends restricting access to adults.

It is also important to avoid suggesting that grief follows a single ‘correct’ process, or that maintaining a connection with someone who has died is itself unhealthy. Child Bereavement UK recognises that people grieve in different ways and at different paces.

If a child is already using a griefbot

Ask how they came across it, what they talk about, and how the conversations leave them feeling. It can be useful to understand whether they mainly revisit familiar stories or seek new advice, reassurance, and approval from the simulation.

Avoid dismissing their feelings or suggesting that they have been silly or made a foolish choice. A child can understand that the responses are generated and still find the experience difficult to give up.

Discuss specific answers where necessary, especially if the child is treating them as reliable statements of what the deceased person believed or wanted. Explain that even a convincing imitation can invent information and give inappropriate advice.

SAIFCA recommends moving away from griefbot use, with trusted people available to support the child. Where a strong dependency has formed, seek professional advice about managing this sensitively and don’t treat it as a disciplinary issue. Stop an interaction that presents an immediate safety concern, and seek urgent help if the child is at risk.

You can offer opportunities to remember the person through stories, photographs, original recordings, familiar activities, or family traditions, according to the child’s wishes.

Review any personal information shared with the provider and the available deletion options.

What schools can do

A school’s response should sit within its existing bereavement, pastoral, and safeguarding arrangements. SAIFCA recommends that schools:

Do not create or recommend simulations of deceased relatives or friends for pupils.
Include griefbots in staff awareness of AI risks, particularly for those supporting bereaved pupils.
Respond without blame if a pupil discloses use, involving the designated safeguarding lead where there is harmful content, dependency, serious distress, or another safeguarding concern.
Ensure pupils know which adults they can approach and how to ask for help.

Educational recreations of historical figures require their own assessment of accuracy, purpose, age appropriateness, and design. Their educational label should not exempt them from scrutiny, however they fall outside this article’s focus on personal bereavement.

What needs to happen next?

Families and schools should not have to resolve these issues after a child has formed an attachment. SAIFCA recommends that:

Griefbot services should not be offered to under-18s, including through features within general-purpose AI products and AI companions.
Bereavement professionals, ethicists, children’s rights specialists, regulators, and standards bodies should examine consent, posthumous identity, family interests, and psychological risks before release.
Providers should be accountable for exploiting grief through advertising, payment pressure, or design that encourages emotional dependency.
Independent scrutiny should address how services behave over repeated conversations, how personal information is handled, and what happens when access changes or ends.
Children should never be treated as a population on which to test commercial products, especially when the risks remain poorly understood.

Bereaved families deserve compassionate support and clear information about what is being offered to them. A convincing imitation of someone they love places particular responsibilities on its creators, especially where children are involved.

Bereavement support and resources

The following organisations provide support and helpful information:

Child Bereavement UK - Support for bereaved children, young people, and families, alongside guidance and training for adults and professionals.

Dougy Center - Bereavement information and resources for children, teenagers, parents, carers, and people supporting grieving families.

Child Helpline International - International directory. Find child helplines by country for emotional support and other concerns. These are general child-support services rather than specialist bereavement services, and availability varies by location.

A local healthcare professional or school support team can also help identify appropriate bereavement services.

References and further reading

Hollanek, T., and Nowaczyk-Basińska, K. (2024). Griefbots, Deadbots, Postmortem Avatars: On Responsible Applications of Generative AI in the Digital Afterlife Industry. Published in Philosophy & Technology. Read the paper.
SAIFCA. AI Companion Chatbots: The Risks to Children. Read the guidance.