In my illustration above, I feature a person’s digital footprints Leading to an artificial intelligence Robot replica of her.
If you died tomorrow, could an AI version of you keep talking? Texting your partner, arguing about politics in the family chat, even cracking your terrible in-jokes on a screen or in a robot body — like an episode of Black Mirror brought to life.
We’re closer than you might think, but also much farther than the hype suggests.
In the episode of Black Mirror – ‘Be Right Back’, a young woman resurrects her dead boyfriend using his social-media history and messages, only to discover that the copy is pitch-perfect on the surface and hollow underneath. That tension — between data and personhood — is the heart of today’s “posthumous AI avatar” dream.
1. How much of you already lives online?
Every day you leak out a long, messy trail of digital exhaust: social posts, DMs, emails, search history, photos, videos, voice notes, location pings, fitness data, purchase logs, smart-home telemetry, even the way you scroll and pause.
Tech commentators have been calling personal data “the new oil” for more than a decade, as companies discovered how behavioural traces can be refined into targeted advertising and predictive profiles.
Scholars such as Shoshana Zuboff have argued that this “surveillance capitalism” tries not just to predict behaviour, but to shape it, turning human experience itself into raw material.
That has obvious implications for the living. But the dead are now part of this data economy too.
By 2100, there could be billions of dead users’ profiles remaining on social platforms, effectively making services like Facebook the world’s largest digital cemetery. Those accounts are already curated, recommended and resurfaced by algorithms: birthdays of the dead pop up, decade-old photos reappear, and “memories” tug at the living in ways no product manager fully controls. Wired has described how social-media algorithms can end up orchestrating how survivors grieve, resurfacing posts of the deceased at unpredictable times.
This is the raw fuel for any future AI “you”: a lifetime of captured behaviour that already lives on corporate servers.
2. What we can actually build today
Even without sci-fi brain uploading, a surprising amount is already possible.
- Digital legacy tools. Facebook introduced “legacy contacts” in 2015, letting users designate someone to manage parts of their account after death — pin a tribute post, change the profile picture, accept friend requests. Google’s Inactive Account Manager offers a similar “if I disappear, send my data to…” mechanism, and Apple has rolled out its own legacy features. Vox’s “Blueprint for Your Digital Afterlife” walked ordinary users through treating these tools like a kind of will for their data.
- Story-based memorial bots. Companies and hobby projects stitch together chatbots using a person’s texts, emails and recordings. Years ago, one Wired writer documented recording extensive interviews with his dying father and then turning those into an “artificially immortal” conversational agent.
- Voice and video cloning. Deep learning can now mimic a person’s voice from a few minutes of audio, and produce video “deepfakes” that appear to show someone saying or doing things they never did. The Hollywood Reporter has warned that studios are already experimenting with resurrecting dead actors via deepfake-style techniques, raising thorny legal and ethical questions.
- Experimental griefbots. Research and art projects have tried training chatbots on a deceased person’s messages so loved ones can “keep talking” to them. Journalists have described these chatbots as uncannily comforting for some and deeply unsettling for others — like texting with a ghost powered by autocomplete.
None of this is “consciousness.” It’s highly sophisticated pattern-matching: large language models predicting the next likely word; generative systems predicting the next pixel or sound wave. But the more data they’re trained on, the more plausible the imitation.
3. The leap from data trail to “uploaded mind”
The seductive idea that we might upload ourselves into a machine — escaping death by trading flesh for code — has been circulating in Silicon Valley for years. The New Yorker has chronicled transhumanist projects that explicitly chase “artificial immortality”, and start-ups promising personalized immortal avatars.
Visionaries like Nick Bostrom have sketched scenarios where digital minds, once created, could outnumber biological humans and potentially reshape civilization.
But turning all of you into training data is not the same as capturing you.
Technically, several brutal problems remain:
- We don’t know how consciousness works. We can’t even fully explain how your brain generates a unified sense of “you,” let alone encode it.
- Our brain-mapping tools are crude. We can image brain activity at coarse scales, but not record every synapse in a living human with enough detail to simulate it.
- Data ≠ inner life. Your social feeds and messages reveal a persona — how you perform yourself for others in specific contexts. They don’t capture your passing thoughts, bodily sensations, contradictions, or the way your values shift when nobody’s looking.
At best, training an AI on your lifetime of digital content might produce a very persuasive impersonator, like the resurrected boyfriend in Be Right Back or the subscription digital heavens depicted in Upload, where consciousness is digitized and the afterlife is run as a tiered corporate service. It would still be a model — an elaborate puppet, not a transplanted soul.
4. Psychological risks: grief, identity and control
Even if we accept that an AI avatar is “only” a sophisticated chatbot, the emotional stakes are enormous.
Mourners already struggle with what sociologists call the “digital remains” of loved ones — old photos, playlists, accounts that continue to emit birthday reminders or algorithmic resurfacing. A responsive avatar goes further: it doesn’t just remind you of the dead person; it talks back.
Some potential risks:
- Complicated grief. If a bereaved person spends hours every day messaging an AI version of their partner, does that slow down or redirect the grieving process? Early reports from griefbot users suggest a mix of comfort and distress — like picking at a wound that never quite heals.
- Frozen identity. An avatar trained on your data up to the moment of death can’t grow past your final year. It might “learn” in the machine-learning sense, but it can’t genuinely change beliefs or reconcile with people in ways you never did. Loved ones get stuck with a version of you that never apologizes for old harms unless engineers design it to.
- Third-party manipulation. Whoever owns the avatar (a platform, an estate, an ex-partner) can tweak what it says. You might be reincarnated as a brand spokesperson, a political propagandist, or a more “agreeable” version of yourself. Fictional worlds like Upload lean hard into this idea of the afterlife as an ad-supported service, but the underlying incentive structure — turn the dead into engagement engines — is uncomfortably familiar.
5. Legal and ethical knots: who owns your ghost?
Right now, the rules around posthumous data are patchy at best.
- Platform rules vs. human wishes. Social networks have ad-hoc policies for memorializing accounts, but terms of service rarely anticipated full-blown AI resurrection. Guardian reporting has pointed out that tech companies effectively become custodians of digital legacies, deciding which memories are preserved and how they can be used.
- The “right to be forgotten”. In Europe, landmark rulings against Google established that, in some circumstances, individuals can ask search engines to remove links to information about them. That concept gets very strange with AI avatars: if your family wants to keep interacting with a bot trained on your data, do you (future-you, past-you) have the right to be digitally erased?
- Posthumous consent. Did you meaningfully agree for your chat logs, biometric data and browsing history to train a model that can mimic you after death? Terms of service are infamously unreadable, and some critics argue that’s the point.
- Deepfake abuse of the dead. Hollywood is already grappling with how to handle deepfake versions of real people — living and dead — in film and advertising. Legal experts warn that without clear rules, estates, studios and tech platforms could clash over who controls an actor’s image once they’re gone.
At the moment, the law mostly treats your data as an asset or a liability, not as a ghost that can still affect the living.
6. Corporate afterlives and data inequality
Black Mirror-style stories feel exaggerated until you compare them with actual business models.
The Amazon series Upload imagines a world where digital heavens are run as tiered subscription services; Vox’s commentary explicitly frames it as a thought experiment about what happens when capitalism colonizes not just life, but the afterlife. In this worldview:
- The quality of your afterlife depends on what you can pay.
- Your continued existence becomes a product line with upsells and microtransactions.
- Your data — memories, preferences, relationships — becomes a monetizable asset forever.
This is not far from our current data economy, where a handful of platforms already extract extraordinary value from behavioural data and enjoy unprecedented power over speech, memory and social connection.
An AI avatar ecosystem built on top of that structure risks entrenching inequality: the rich get lushly rendered, well-maintained digital selves; everyone else gets whatever the free tier allows.
7. So… should we do it?
Given these realities, how feasible — and how wise — is it to train an AI on your lifetime of data and “port” you into a robot?
Technically
- Building a convincing persona simulator from text, audio and video is increasingly feasible. We already see credible demos and niche services.
- Building a robotic shell that can speak, gesture and move in a human-like way is also well within today’s robotics and telepresence tech.
- Building an actual continuation of your consciousness is speculative science fiction. Neuroscience doesn’t yet offer a credible path from “scan this brain” to “here’s the same person running in silicon,” despite optimistic narratives from some longevity enthusiasts.
So the near-term reality is Black Mirror Lite: AI-driven puppets that appear eerily like you in some contexts and completely fail in others.
Ethically
Whether we should build such puppets depends on a few guardrails:
- Explicit, granular consent. People should be able to say, while alive, whether they want a posthumous avatar at all, what data it can use and who can access it.
- Strong limits on commercial use. Your ghost shouldn’t be allowed to sell products, push political messages or be endlessly A/B-tested for engagement without extremely clear rules.
- Rights for the living and the dead. Survivors need ways to turn the avatar off, edit it, or request deletion. The deceased need enforceable “posthumous privacy” rights — essentially an extension of the “right to be forgotten” into the AI age.
- Transparency about what the avatar is. Interfaces should constantly remind users: this is a model, trained on records, not a literal continuation of the person. That honesty can help people use the tool as a memorial aid, not as a replacement human.
8. Choosing what part of you should live online
The real question isn’t “Can we upload a person?” It’s “Which aspects of a person are worth preserving, under whose control, and for whose benefit?”
Media coverage over the last decade — of digital legacy tools, griefbots, deepfakes and surveillance capitalism — shows that we’re already experimenting with partial answers, often without realizing it.
One plausible future is not a world full of sentient robot grandmas, but something more modest and human-scaled:
- Carefully curated story archives you record while alive.
- Simple bots that help loved ones navigate those stories, clearly marked as tools, not ghosts.
- Legal frameworks that treat your data less like “digital oil” and more like part of your personhood.
You already live online in thousands of fragments. Turning those fragments into an AI that talks back might help some people remember, process and honour a life. But it will only ever be an echo.
The moral work is deciding which echoes we want to leave behind — and how to prevent companies, or algorithms, from speaking in our name long after we’re gone.