Posthumous AI representations of the dead are now a reality rather than a future possibility, and at collective scale they raise governance questions that societies are only beginning to confront. These representations range from the modest — a chatbot trained on a person's text messages, capable of responding in their style — to the substantial — high-fidelity audiovisual avatars capable of extended interactive conversation, trained on years of recorded speech and documented preferences. In each case, the representation produces content the deceased never created, in a voice and style the deceased would recognize as their own. This is a form of posthumous speech, and at collective scale it is also a form of collective memory manipulation with no clear precedent in the history of human civilization.
The distinction between digital immortality (article 6464) and posthumous AI representation is primarily one of frame and emphasis: where digital immortality foregrounds the longing of the living for continued presence, posthumous AI representation foregrounds the social and political reality of what is produced — a simulacrum that makes claims, expresses apparent preferences and values, and participates in social life in ways that can influence the living. This distinction matters because the governance challenges are different. The grief management question is a healthcare and cultural question; the posthumous representation question is a legal, political, and epistemic question. A posthumous AI of a public figure that produces apparently authentic statements on political, social, or commercial topics is not merely a grief tool; it is a new category of speech act with consequences in public discourse, reputation, and collective memory.
Law 5 — Revise — is the primary law because the social institutions tasked with managing posthumous representation — law, ethics, journalism, memory institutions, religious traditions — were designed for a world in which the dead stop speaking. They are now under pressure to evolve in response to a world in which the dead can continue to speak, indefinitely, with increasing fidelity, and in new domains they never addressed. The evolutionary challenge is compounded by the fact that the technology is developing faster than institutional capacity to understand and respond to it, and by the fact that the commercial interests driving the technology have incentives that do not align with the social interests that the institutional evolution is meant to serve.
Law 0 — the law of orientation and foundational observation — establishes the epistemic ground from which all institutional response must depart. Before societies can design governance for posthumous AI representations, they must be clear about what those representations actually are: not the person, not a continuation of the person's consciousness or experience, but a statistical model of the person's communicative patterns trained on digital traces they left behind. This is not a minor or merely semantic point. The tendency to attribute authentic personhood to sophisticated simulations — the ELIZA effect, well-documented since Joseph Weizenbaum's 1960s chatbot demonstrated that people formed parasocial bonds with very simple programs — means that without deliberate epistemic vigilance, individuals and institutions will systematically overestimate the authenticity of what posthumous AI representations express. Governance frameworks that mistake the representation for the person will reach wrong conclusions about consent, accountability, and rights.
Law 3 — the law of relation and community — is the secondary law because posthumous AI representations do not exist in isolation from the communities that produced the deceased and that now must manage their digital continuation. The deceased was embedded in networks of relationship, obligation, and meaning. The representation continues to participate in those networks while being constituted by only a partial and static record of the person's communicative life — the digitally captured subset, which is typically skewed toward formal or public communication rather than the full richness of embodied presence. The relational communities of the deceased — family, friends, professional networks, cultural communities — have legitimate interests in how the person is represented posthumously, and these interests may conflict with each other, with the interests of companies holding the data, and with whatever interests we attribute to the deceased themselves. At collective scale, communities of the dead's survivors constitute a new kind of stakeholder group whose interests require explicit institutional recognition.
The consent architecture of posthumous AI representation is in crisis. The data required for training — texts, emails, social media posts, recorded calls, video, documents — was generated by the deceased in contexts that did not include posthumous AI training as a contemplated use. In most jurisdictions, this data belongs legally to the platforms that stored it, not to the individual who generated it. The individual's family may have access to some of it through inheritance or platform legacy policies, but the legal picture is fragmented, inconsistent across jurisdictions, and far from providing a coherent consent framework. Beyond the data, the representation itself — the AI model trained on the data — is a new kind of artifact whose ownership and control are entirely unclear. Is it a portrait? A biography? A product? A person? The category determines the applicable law, and no jurisdiction has clearly answered the question.
The political and epistemic risks of posthumous AI representations at collective scale extend beyond individual grief management to the integrity of collective memory and public discourse. If high-profile figures — political leaders, intellectuals, religious authorities, artists — can have AI representations created that produce apparently authentic new statements, the potential for posthumous reputation manipulation is substantial. The deceased cannot correct a misrepresentation, cannot context their statements, cannot repudiate a fabrication. The survivors who might contest such representations face information asymmetries and legal vacuum. In authoritarian contexts, the potential for state manipulation of posthumous representations to serve current political narratives is particularly acute. The broader epistemic problem is that in an information environment already saturated with synthetic content, posthumous AI representations of well-known figures add another vector of plausible-seeming content that degrades the epistemic infrastructure of public discourse.
What Law 5 demands here is not the suppression of posthumous AI representation — the technology exists, the commercial demand exists, and many uses are genuinely benign or beneficial — but the development of governance frameworks sophisticated enough to distinguish between use cases by their social functions and consequences, and to apply appropriate regulatory, ethical, and institutional responses to each. This requires the coordinated evolution of multiple institutional domains: law must develop posthumous digital personhood frameworks; journalism and media must develop authentication standards for posthumous AI content; memory institutions must develop archival practices that preserve context alongside representation; communities must develop social norms that constrain exploitative or harmful uses. The evolutionary challenge of Law 5 is precisely this multi-domain coordination across institutions with different incentive structures, different timescales, and different relationships to the technology.