By
Stuart Kerr, Technology Correspondent, LiveAIWire

A South Korean documentary broadcast in 2020 showed a mother
interacting with a real-time digital reconstruction of her seven-year-old
daughter, who had died of a rare illness three years earlier. The
reconstruction moved, spoke, and responded using video footage, voice
recordings, and AI synthesis. The mother wept and said it felt like a
miracle. Ethicists called it a preview of one of the most contested questions
in artificial intelligence: what rights, if any, do the dead have over their
own digital likeness?

The technology enabling digital posthumous reconstruction  — 
generative AI trained on voice, video, written communication, and
social media data  —  has matured significantly since that
broadcast. What began as high-cost, studio-produced projects for
entertainment and documentary purposes is becoming a consumer-accessible
capability. The ethical framework governing it has not kept
pace.

What the Technology Can Now Do

Current AI systems can clone a voice from as little as three
seconds of audio, generating new speech in that voice from any text input.
Video synthesis systems trained on video footage can reconstruct a person’s
appearance, facial expressions, and characteristic gestures with high
fidelity. Large language models fine-tuned on a person’s writing  — 
emails, social media posts, diary entries  — 
can generate new responses that reflect their characteristic
vocabulary and reasoning patterns.

Several commercial services now offer posthumous digital companion
creation as a product. StoryFile and HereAfterAI allow individuals to record
their responses to hundreds of questions while alive, creating an interactive
AI that family members can query after their death. A smaller number of companies
offer reconstruction services for people who did not prepare in advance,
working from available data with the consent 
—  often contested  — 
of surviving relatives.

What this means for you: the digital data you generate throughout
your life  —  messages, voice recordings, video calls,
social media posts  —  constitutes a training dataset from which a
synthetic version of you could be constructed, with or without your prior
consent, after your death. The legal frameworks that would govern this
reconstruction are embryonic in most jurisdictions.

The Grief Argument and Its Limits

Proponents of digital resurrection technology frame it primarily
as a grief support tool. The bereaved find comfort in continued interaction
with a digital representation of someone they have lost; this comfort is real
and measurable. Grief counsellors have noted that some clients find AI
companions based on deceased relatives helpful in the immediate aftermath of
loss, particularly when the death was sudden and unresolved conversations
were left incomplete.

The psychological risks are also real. The same feature that makes
a digital reconstruction comforting 
—  its responsiveness and
apparent continuity  —  may delay rather than support the grieving
process. Dependency on a synthetic version of the deceased may prevent the
kind of acceptance that grief research consistently identifies as necessary
for long-term psychological adjustment. Research published in the journal
Death
Studies has examined the psychological dimensions of continuing
bonds with the deceased, noting that technology-mediated bonds require
careful clinical attention to avoid pathological grief
outcomes.

Consent, Data Rights, and the Dead

Most legal systems do not extend personality rights or data rights
to the deceased. A person’s voice recording, video footage, and written
communication typically pass to their estate, and surviving relatives or
executors have broad latitude over how that material is used. The gap between
this legal reality and the moral intuitions of most people  — 
who believe that the deceased would have strong preferences about
whether an AI is trained to speak in their voice  — 
is the central ethical tension in digital
resurrection.

Several jurisdictions are beginning to address this. California
passed legislation in 2023 requiring explicit consent for the use of a
deceased performer’s digital likeness in AI-generated commercial productions.
The EU’s AI Act contains provisions relevant to synthetic media involving
real individuals. Neither framework comprehensively addresses the consumer
use case  —  a family member using a commercial service
to reconstruct a deceased relative without their prior consent.

The question of who speaks for the dead in matters of digital
identity connects to broader debates about what
constitutes authentic selfhood in an era of synthetic
representation.

Commercial and Reputational Risks

Digital necromancy creates obvious opportunities for misuse. A
deceased public figure’s voice or likeness reconstructed to endorse a
product, express a political view they never held, or make statements
damaging to their estate could cause significant harm to their reputation and
distress to their relatives. The technology to produce such content is
increasingly accessible; the legal remedies for the harms it causes are
unclear.

The entertainment industry has confronted a version of this
question through the posthumous use of deceased actors and musicians in new
productions. The use of a digitally reconstructed Peter Cushing in a Star
Wars film, and of AI-generated voices for deceased musicians in commercial
releases, generated significant public debate about the rights of the dead
and the commercial interests of estates. Reporting
on AI voice cloning in music has highlighted the scale and
accessibility of the technology and the inadequacy of current
protections.

A Framework That Does Not Yet Exist

What is needed is a legal and ethical framework that distinguishes
between different uses  —  private grief support, commercial
exploitation, political manipulation 
—  and applies proportionate
standards of consent and accountability to each. Such a framework would
require legislation, platform governance, and the development of cultural
norms around what constitutes respectful treatment of the digital
dead.

The absence of that framework does not prevent the technology from
advancing. It means that the harms it enables will continue to outpace the
remedies available to those affected. The debate around algorithmic
governance of digital content is directly relevant: who decides
what a deceased person’s digital representation can and cannot be used for,
and who enforces those decisions, are questions that platforms, legislators,
and families will all have to answer.

The absence of consent frameworks for posthumous digital use
mirrors problems identified in AI
systems that encode historical biases without the knowledge or
agreement of those whose data shaped the model. In both cases, the harm
accrues to people who had no meaningful opportunity to
refuse.

The industry developing digital resurrection
services exists in a regulatory vacuum that is unlikely to persist. Several
jurisdictions are actively developing legislation, and international
coordination on synthetic media governance is advancing through UNESCO and
the Council of Europe. The question is not whether governance will arrive,
but whether it will arrive before the commercial market has normalised
practices  —  including non-consensual reconstruction and
commercial exploitation of deceased individuals’ likenesses  — 
that will then be much harder to prohibit. The history of social media
regulation suggests that market normalisation consistently outpaces
legislative response, and that the harms generated in the interim are borne
by the most vulnerable individuals rather than the most commercially powerful
platforms.

The social norms around digital resurrection are
forming now, in the absence of adequate legal frameworks, and they are being
shaped primarily by the commercial interests of the companies providing the
technology. That is a pattern familiar from the early development of social
media, where norms of disclosure, privacy, and consent were established by
platform design choices rather than democratic deliberation. The consequences
of those early norm-setting decisions have proven difficult to reverse. The
same risk applies to digital necromancy: the practices that become normalised
in the next five years will be much harder to prohibit or reform in the
decade after.

About the Author

Stuart
Kerr is a technology correspondent at LiveAIWire, covering artificial
intelligence, emerging technologies, and their impact on society and
industry.