At CES 2026 in Las Vegas, a software company called MyPersonas demonstrated technology that can build a digital replica of an employee from their voice recordings, video footage, and accumulated knowledge base. The replica can attend virtual meetings, answer questions in the employee’s communication style, and represent their professional persona when the original is unavailable.

The demo triggered immediate headlines about digital immortality. Several commentators framed it as a step toward uploading workers into the cloud. The more honest framing is: this is sophisticated behavioral modeling, not consciousness transfer. But the gap between what MyPersonas actually does and what it implies about the trajectory of digital identity technology is precisely where the interesting questions live.

What MyPersonas Actually Does

MyPersonas is a multimodal machine learning system. It ingests training data from three sources: audio samples of the target person’s voice, video recordings capturing facial expressions and communication patterns, and a structured knowledge base derived from documents, emails, or expertise databases the person has contributed to.

From this, it builds a generative model. The voice component produces speech in the person’s acoustic signature. The video component generates corresponding facial and gestural animation. The knowledge component provides a retrieval-augmented generation layer, meaning the digital twin can answer domain-specific questions using information the real person contributed.

The result is a behavioral approximation. The twin sounds like the person, moves like them, and answers questions in a way consistent with their documented expertise. It does not have experiences, does not develop new knowledge, and does not continue the person’s psychological development. When new information is not in the training data, the system either generates a plausible-sounding answer or acknowledges uncertainty.

This is TRL 6 to 7. The technology is demonstrated in commercial products. It works within its defined scope. What it produces is not a person. It is a very detailed mask.

Why Employers Want This

The commercial pitch is straightforward. Knowledge workers accumulate expertise, relationships, and communication patterns over years. When they leave a company, retire, or become unavailable, that accumulated value is partially lost. MyPersonas offers a way to retain access to it.

The demo scenarios at CES were corporate: a senior executive who can’t attend a client meeting sends their digital twin instead. A retiring engineer’s knowledge is preserved for onboarding new hires. A consultant working across time zones delegates routine interactions to their replica while focusing on higher-complexity work.

These are plausible near-term use cases, and they explain why enterprise software companies are interested. The business model treats a person’s professional self as an asset that can be extracted, stored, and deployed independently of the person’s time and availability.

The immortality framing comes from extending this logic past death. If a company can use an employee’s digital twin while they’re on vacation, the same infrastructure could, in principle, run the twin after the employee dies. The digital employee continues attending meetings, answering questions, representing the organization’s accumulated knowledge.

The commercial grief tech industry has already built smaller-scale versions of this for personal use: companies that build conversational AI personas from a deceased person’s texts, emails, and social media posts. MyPersonas is the enterprise equivalent, optimized for professional knowledge rather than personal relationship.

The Consent Problem

The obvious immediate issue is consent. The CES demo showed systems built from employee data. Employees in most jurisdictions have limited legal protection over how their employers use data they generate at work. Email archives, recorded meetings, project documentation, and performance reviews are typically company property.

MyPersonas builds from this data. An employee who has been generating professional content for a company for fifteen years may find that their employer has constructed a behavioral model of them without explicit consent for that specific use. When the employee leaves, the company may retain the model. When the employee dies, the company may continue deploying it.

The ethics of posthumous digital identity apply here with specific force. A personal digital ghost created for grief processing operates in a private relational context. An enterprise digital twin operates in commercial contexts, potentially for profit, representing the person’s professional reputation to clients and colleagues. The stakes are higher and the potential for misuse is broader.

Current employment law does not clearly address this. Data protection frameworks like GDPR give individuals rights over personal data but carve out substantial exceptions for employment contexts. The legal landscape for digital identity assets in the workplace is largely unwritten.

The Gap Between Behavioral Twin and Mind Upload

The MyPersonas demo reveals something important about the distance between current digital cloning technology and genuine whole brain emulation. The gap is not primarily technical. It is conceptual.

A behavioral twin captures outputs: how a person speaks, what they know, how they respond to prompts within a defined domain. It models the interface between a person and their environment, specifically the professional interface.

Whole brain emulation aims at something categorically different: preserving the underlying neural processes that generate behavior, including aspects of cognition that are never externalized in professional communication. Internal monologue, emotional processing, aesthetic experience, ambiguous uncertainty, the background hum of a person’s ongoing mental life. None of this is captured in a training set of meeting recordings and documentation.

A digital twin might accurately predict how its source would respond to a familiar question. It would fail to represent how that person would respond to genuinely novel situations, because novelty, by definition, is not in the training data. A genuine mind upload, if such a thing were possible, would carry the source person’s capacity for new thought, not just a record of past thought.

Digital doppelgangers raise these questions directly. Is a replica that accurately represents who someone was sufficient to preserve them? If the answer depends on what preservation is for, different use cases give different answers. For corporate knowledge retention, behavioral accuracy may be enough. For personal identity continuity, it is almost certainly not.

What Cloud Deployment Actually Means

The framing of digital immortality through cloud-hosted personas is worth examining carefully. Distributed cloud consciousness raises genuine questions about what it would mean to exist in a server environment. MyPersonas is a long way from that. But the language used to describe it, “digital twin,” “virtual employee,” “preserved persona,” borrows conceptual vocabulary from mind uploading discourse without engaging with its implications.

When companies describe a deployed digital twin as the employee “continuing to work,” they are making an identity claim. They are asserting, even implicitly, that the replica represents the person’s presence in the organization. This is a philosophical position, not just a marketing choice. It carries assumptions about what a person is and what counts as their participation in an activity.

Those assumptions deserve to be examined. Behavioral replication at high fidelity is a technically impressive achievement. Calling it immortality, or even continuity, requires more justification than the technology currently provides.

Where This Is Going

The enterprise digital twin market will grow. MyPersonas is one of several companies building in this space. As large language models improve and multimodal training becomes cheaper, behavioral digital twins will become more accurate, more context-aware, and more convincing.

The near-term developments to watch are: dynamic knowledge updating (twins that can learn new information after training), affective modeling (twins that capture emotional range, not just professional communication patterns), and longitudinal consistency (twins that maintain coherent development over time rather than being frozen at a training snapshot).

Each of these moves the technology closer to what would count as genuine identity continuity, and further into philosophical territory that the software companies building it have not publicly engaged with. The consent frameworks, identity rights, and deployment ethics will need to be worked out by regulators, employers, and individuals before the technology makes those decisions by default.

What CES 2026 demonstrated was not digital immortality. It was the infrastructure that will make the question unavoidable.

---

Official Sources

• MyPersonas AI, CES 2026 demonstration — Euronews report (January 2026): euronews.com
• General Data Protection Regulation (GDPR), Regulation (EU) 2016/679 — employment data exceptions, Article 88
• Tene, O., & Polonetsky, J. (2013). “Big Data for All: Privacy and User Control in the Age of Analytics.” Northwestern Journal of Technology and Intellectual Property, 11(5)
• Floridi, L., et al. (2018). “An Ethical Framework for a Good AI Society.” Minds and Machines, 28(4), 689-707