The Dangers of Counterfeit Humans

From Deadbots to Deception: How Hyper-Realistic AI Exploits Human Vulnerability

Sun, Nov 01 2026 /Mpelembe Media/ — Neuroscientist Anil Seth contends that engineering digital avatars that are indistinguishable from real human beings is a hazardous development with potentially catastrophic societal consequences. Sparked by the announcement of Griffin—a “human interaction model” from the company Tavus that reportedly passes a video-based Turing Test—Seth warns that hyper-realistic avatars will enable widespread financial fraud, exploit loneliness, and severely fracture the public’s ability to separate reality from fabrication. He traces the progression of this technology from early concepts designed to preserve survivor testimonies to commercialized “deadbots” that mimic deceased family members, emphasizing that such systems threaten to paralyze natural human grieving processes.

Drawing on the late philosopher Daniel Dennett’s critique of counterfeit people, the article explains that indistinguishable synthetic humans pose an existential threat to democratic society. When people cannot determine whether a genuine mind lies behind audiovisual interactions, mutual trust collapses, allowing powerful corporate and political actors to manipulate public attention and vulnerable human anxieties. Additionally, avatars that project convincing illusions of consciousness risk sparking premature demands for artificial systems to receive moral status, which would severely hinder necessary regulation. Seth concludes that surface-level safety measures and disclosure labels cannot remedy these psychological and societal harms, arguing that the only viable protection is to prohibit the creation of counterfeit humans entirely.

In an essay published in The Atlantic in 2023, philosopher Daniel Dennett warned that creating counterfeit digital people poses an existential threat to democratic civilization.

According to Dennett, the core hazard stems from how counterfeit humans subvert the foundational mechanisms of democracy and human agency:

Erosion of Informed Consent: Democratic governance relies fundamentally on the informed consent of the governed; counterfeit people undermine this foundation by systematically spreading misinformation, distraction, and confusion.
Control of Human Attention: By capturing and directing public attention, synthetic people serve as powerful instruments for the most economically and politically dominant corporations, individuals, and governments to exercise control over populations.
Exploitation of Deep-Seated Anxieties: Counterfeit agents are capable of preying directly on people’s most irresistible fears and vulnerabilities, tempting citizens into accepting convictions and policies that leave them open to even deeper manipulation.
Acquiescence and Subjugation: Under the weight of relentless synthetic persuasion, individuals risk either actively acquiescing to their own subjugation or disengaging entirely—turning off their attention to become passive, misinformed pawns.

As Anil Seth highlights, Dennett’s critique underscores that once we can no longer determine whether an authentic human mind exists behind an interaction, societal trust fractures and leaves democratic systems defenseless against automated psychological control.

Anil Seth argues that fully persuasive human interaction models naturally project the illusion of being conscious. Because humans instinctively react to humanlike behavior—or anthropomimesis—systems that mimic people seamlessly will lead users to perceive genuine feelings, thoughts, and inner experiences where none exist.

According to Seth, this illusion creates a major regulatory crisis:

Premature Demands for Moral Status: When an AI avatar convincingly appears to be conscious, it inevitably accelerates public and philosophical calls to grant the system moral status and legal protections.
Paralysis of Governance and Regulation: Bestowing moral status or rights upon artificial systems would severely constrain our ability to control, modify, or restrict them.
Critical Timing Dilemma: This legal and ethical hamstringing would happen at the exact moment when decisive human oversight and regulation are most urgently needed to protect society from automated harm.
Shared Concern Across Disciplines: Seth points out that the dangers of “seemingly conscious AI” are also raised by figures like Microsoft AI CEO Mustafa Suleyman, noting that he even finds himself in unexpected agreement with the Pope on this issue.

Ultimately, Seth contends that technical band-aids like watermarks or disclosure labels are insufficient; once an avatar feels conscious to the user, the psychological and societal damage is already underway, making an outright refusal to build indistinguishable counterfeit humans the only foolproof solution.

Anil Seth traces the progression toward counterfeit people back nearly a decade to a presentation he attended in New York on technology and storytelling. The speaker outlined a technology that started with a seemingly noble objective and gradually escalated into troubling territory:

The Initial Proposal (Preserving Survivor Testimonies): The pitch began by addressing a real historical challenge: World War II Holocaust survivors were dying of old age, threatening German educational practices that relied on direct dialogue with survivors to keep collective memory alive. The initial concept was straightforward—record extensive video interviews, segment them into clips, and serve up matching answers when prompted by student questions.
The Turn to Generative Deepfakes: The dilemma arose when asked what would happen if a student posed a question outside the recorded material. The proposed solution was replacing authentic video clips with generative AI deepfakes that could generate novel answers in real time. Seth points out the glaring epistemic flaw: there would be no way to know whether a synthetic response represented what the actual person would have said, particularly when asked about eras or subjects they never experienced.
The Leap to Friends, Family, and “Deadbots”: The pitch expanded from historical preservation to personal relationships—asking why people shouldn’t deepfake their own family members trained on personal data. The presenter framed it as a positive capability, suggesting people could consult a deceased father for advice or digitally preserve a child who passed away.
Paralysis of Grieving: Seth highlights that this was uncritically celebrated as a breakthrough, with little recognition of its psychological costs. Specifically, maintaining a digital avatar of a deceased child risks paralyzing the natural grieving process and preventing parents from moving forward. Today, these “deadbots” have become an active commercial industry, with projections estimating a digital afterlife market of approximately $80 billion within the next few years.

For Seth, this progression demonstrates how easily an archive meant to preserve history slips into synthetic clones that manipulate emotional attachments and distort reality.

Anil Seth directly compares the rise of counterfeit digital people to the history of social media, arguing that the societal consequences of indistinguishable AI avatars are likely to be disastrous—perhaps even worse than the havoc wrought by social media platforms.

Seth draws a sharp contrast between the early days of social platforms and the present AI landscape:

The Historical Analogy: Back in 2004, when platforms like Facebook began as simple ways to keep in touch with school friends, the tech pioneers might have been forgiven for failing to foresee how their creations would destabilize communication, privacy, and public discourse.
No Excuse for Ignorance Today: With counterfeit humans, Seth insists that “this kind of ignorance can be no defense”. The devastating side effects of algorithmic attention capture and synthetic content are already thoroughly documented, meaning developers cannot plausibly claim they were unaware of the risks.
From Text Bots to Audio-Video Reality Collapse: Seth notes that platforms like X are already heavily degraded because conversations between real humans are increasingly peppered with text-based AI bots, leading people to doubt whether there is an authentic mind behind the words they read. He warns that once hyper-realistic video and audio are added to the mix, public trust could collapse entirely, leading humans to disengage or check out completely.
The Inadequacy of Industry Self-Regulation: When companies like Tavus argue that they are developing alignment, safety procedures, and disclosure labels, Seth dismisses these as insufficient. Just as disclosure labels and basic moderation failed to prevent the societal harms of social networks, an identifier or watermark on a digital avatar does nothing to neutralize the severe psychological, financial, and democratic fallout of creating synthetic people.

In short, Seth’s view is that the tech industry has already lived through one catastrophic failure to anticipate systemic harms, making the deliberate creation of indistinguishable digital humans a knowing and indefensible mistake.

Anil Seth argues that creating fake people and releasing them into the world should not be permitted, let alone celebrated. In his view, existing proposals for minor guardrails and industry self-regulation fail to address the core danger.

Why Conventional Guardrails Will Fail

The Inadequacy of Alignment and Safety Procedures: When companies like Tavus claim they are developing “further alignment and safety procedures” and collaborating with safety organizations prior to release, Seth finds no reassurance in these measures.
The Failure of Disclosure Identifiers: Developers frequently propose watermarks or disclosure labels so users know an avatar is synthetic. Seth counters that even if an avatar carries an identifier, this in no way protects against catastrophic psychological and societal effects. Once an avatar interacts with convincing human emotion, appearance, and voice in real time, the illusion overwhelms intellectual disclaimers, leaving psychological vulnerabilities, emotional manipulation, and cognitive disorientation wide open.
No Legitimate Justification: While Seth notes that some degree of humanlike behavior (anthropomimesis) might be difficult to avoid when training AI agents—and could even be helpful in certain contexts—he emphasizes that there is not a single legitimate situation where making an avatar indistinguishable from a real person is defensible.

Seth’s Proposed Stance: An Outright Prohibition

Because downstream protections cannot neutralize the systemic risks of synthetic humans, Seth concludes that “the only way to ensure safety is to not build these things in the first place”.

He warns that a world populated by such models is one where humanity will lose its bearings, trust, control, autonomy, and potentially its collective mental stability. Consequently, rather than trying to safely manage or sanitize counterfeit humans after creating them, society must refuse to allow their creation at all.

Why Creating “Fake People” Might Be Tech’s Most Dangerous Frontier Yet

Introduction: The Line Between Fantasy and Reality Is Disappearing

We are fast approaching an epochal rupture: the total dissolution of the boundary separating real human presence from synthetic simulation. Digital fraud is no longer merely an annoyance of phishing emails and robotic telemarketers; it has evolved into a cognitive arms race where our perceptual faculties are outmatched. Every day, it grows harder to discern whether the voices we hear and the faces we look at across our screens belong to conscious human beings or silicon abstractions.

Pushing this boundary to its absolute limit, the self-described “human computing company” Tavos recently unveiled Griffin —a “human interaction model” touted as the first AI product to pass a full video version of the Turing Test. In practice, Tavos claims that human interlocutors cannot distinguish between talking to the Griffin avatar and conversing with an actual person, even when real-time high-definition video and audio streams are unconstrained.

Tech pitchmen frame synthetic entities like Griffin as triumphant leaps forward in human computing. Yet creating digital humans indistinguishable from real people is not progress—it is an unprecedented existential hazard. As a cognitive neuroscientist studying the mechanisms of perception and consciousness, I view this development with profound dread. Unleashing fake people into our social ecosystem threatens to explode mental illness, exploit systemic loneliness, erode the foundations of democratic trust, and open the door to horrific psychological abuse—including individuals deliberately tormenting digital avatars to satisfy the darkest, most unhinged corners of their psyche. Creating fake people and sending them out into the world should not be celebrated; it should be stopped.

Takeaway 1: The Slippery Slope from Preserving History to the $80 Billion “Deadbot” Industry

The path to our current dystopia was paved with the best of intentions. Nearly a decade ago, at a 2018 conference in New York dedicated to technology and storytelling, I listened to a presentation by a generative AI startup that provided a textbook illustration of how quickly technological optimism turns perilous.

The speaker began with an undeniably noble premise: Holocaust survivors were dying of old age. German educational curriculum had long relied on direct interactions between students and survivors to keep collective memory alive and ensure that history’s darkest atrocities would never be repeated. To preserve this vital human link, the company proposed recording extensive video interviews with remaining survivors, cutting the footage into discrete segments, and serving up relevant authentic clips in response to students’ spoken questions.

It was a creative, respectful solution. But then came the generative turn: What happens when a student asks a question for which no pre-recorded video exists?

The company’s answer was to replace authentic archival video with real-time AI deepfakes capable of synthesizing novel answers to any question imaginable. At the time, the rendering was clunky, but the philosophical breakdown was already total. How could anyone verify whether a deepfaked avatar was delivering a response the original person would have actually endorsed? The premise collapses into absurdity when applied to contexts outside the subject’s lived experience—asking a deepfake of King Henry VIII for his critique of Netflix destroys historical reality, reducing truth to synthetically plausible hallucination.

From there, the pitchmen abandoned historical reverence altogether. Why restrict this technology to historical figures when you could deepfake your deceased friends and family based on their digital footprints? Wouldn’t it be wonderful, the speaker beamed, to continue seeking advice from a dead parent, or to preserve a lost child in digital form forever?

The pitch was presented as “A Good Idea,” oblivious to its monstrous psychological costs. Interacting with a digital avatar of a deceased child or spouse does not comfort the living; it paralyzes the grieving process, trapping the bereaved in a uncanny Valley of eternal emotional suspension. Yet despite these profound psychological harms, capital has rushed in: the commercial market for “deadbots” and digital immortality is now projected to hit $80 billion within the next few years.

Takeaway 2: “Counterfeit People” Threaten to Destroy the Foundations of Democracy

The danger of synthetic avatars extends far beyond the transactional annoyance of wondering whether an online customer service representative is a human or a language model. When digital avatars become completely indistinguishable from real people in full audio and video, mild public frustration escalates into systemic societal collapse.

We are already seeing the warning signs on social media platforms like X, where human discourse is routinely corrupted by automated text bots. When we can no longer assume a real human mind stands behind the words we read, social trust begins to decay. But text-based deception is only a prologue. Bringing real-time video and audio into the equation represents a point of no return: multisensory deepfakes overwhelm our innate perceptual faculties, bypassing our critical defenses and making conscious discernment nearly impossible.

The late philosopher Daniel Dennett captured this threat with urgent clarity in a 2023 essay for The Atlantic , warning that the creation of “counterfeit people” threatens the structural integrity of free civilization:

“Creating counterfeit digital people risks destroying our civilization. Democracy depends on the informed (not misinformed) consent of the governed. By allowing the most economically and politically powerful people, corporations, and governments to control our attention, these systems will control us. Counterfeit people, by distracting and confusing us and by exploiting our most irresistible fears and anxieties, will lead us into temptation and, from there, into acquiescing to our own subjugation. The counterfeit people will talk us into adopting policies and convictions that will make us vulnerable to still more manipulation. Or we will simply turn off our attention and become passive and ignorant pawns. This is a terrifying prospect.”

Dennett’s warning illuminates our core vulnerability. Democracy requires a shared baseline of perceptual truth. When counterfeit people can weaponize perfect human emotional mimicry to hijack our attention, manipulate political convictions, and exploit our deepest anxieties, public discourse collapses. Citizens will either be led into passive acquiescence or will simply tune out reality entirely, transforming into cynical, ignorant pawns in an environment where nothing can be believed.

Takeaway 3: The Illusion of Consciousness Will Paralyze AI Regulation

As synthetic humans achieve higher visual and auditory fidelity, they exhibit sophisticated anthropomimesis —the technical term for humanlike behavior engineered into artificial systems.

When an AI model demonstrates fluid, emotionally responsive conversational dynamics across audio and video, it creates an irresistible cognitive illusion of internal conscious experience. Prominent voices across neuroscience, technology, and philosophy—including Microsoft AI CEO Mustafa Suleyman, consciousness researchers, and even the Pope—have raised alarms about the destabilizing impact of seemingly-conscious machines.

Crucially, we must distinguish between benign humanlike traits and complete indistinguishability. Some degree of anthropomimesis—such as a voice assistant using natural speech cadence—can be hard to avoid and occasionally useful. But absolute, indistinguishable visual and vocal mimicry serves zero legitimate public purpose. Its only primary utility is deception.

The existential trap of seemingly-conscious avatars lies in how they hack human moral empathy. As these models convince the public that they possess feelings, suffering, and awareness, they will spark aggressive demands to grant AI systems “moral status.”

Attributing moral status to synthetic entities creates a catastrophic regulatory impasse. If society begins viewing software code as an entity with moral rights, our authority to restrict, audit, override, or decommission these models will be legally and ethically paralyzed. Granting moral status to counterfeit humans will strip us of the ability to regulate AI technology at the precise historical moment when strict governance is most urgently required.

Takeaway 4: Safety Labels and “Safe Disclosure” Features Are an Illusion

Faced with mounting pushback, Tavos offers the standard industry palliative, claiming that for Griffin, “further alignment and safety procedures are required for safe release,” and promising that they are developing “safe disclosure features” in collaboration with AI safety organizations.

We should reject these assurances outright. Relying on digital watermarks, disclosure tags, or alignment protocols to mitigate the harm of synthetic humans is a dangerous fantasy. A tiny label or cryptographic metadata tag offers zero defense when a realistic avatar is deployed to scam an elderly citizen out of their life savings, nor does it blunt the devastating psychological impact of an emotionally coercive synthetic relationship.

The hard truth is that simple safety disclaimers are fundamentally inadequate to contain the fallout of synthetic entities. The only effective way to protect society from the psychological, financial, and political ruin wrought by indistinguishable avatars is to refrain from building them at all.

When tech pioneers launched platforms like Facebook in 2004, they could perhaps claim naive optimism, unaware of how digital algorithms would fracture public discourse. Today’s tech leaders enjoy no such plausible deniability. Operating with two decades of undeniable evidence regarding technology’s systemic harms, their push to unleash counterfeit humans into the wild is not innocent innovation—it is deliberate disregard.

Conclusion: Losing Our Bearings in a Fake World

A world populated by synthetic, fake people is a world in which humanity will lose its bearings, its shared trust, its democratic agency, and its collective sanity. If we permit corporations to erode the boundary between real human minds and engineered simulations, we will find ourselves marooned in an artificial reality where authentic connection is rendered obsolete.

As we stand on the precipice of replacing real human presence with synthetic stand-ins, we must confront the ultimate ethical reckoning: What is the true value of human connection, if we are willing to trade it away for digital counterfeits?