Somewhere inside Meta, engineers are building a digital Mark Zuckerberg. Not a cartoon mascot or a chatbot with his name on it, but a photorealistic, AI-driven 3D avatar designed to talk with the company’s employees, answer their questions, and deliver feedback as though they were sitting across from the CEO himself.
The project, first reported by the Financial Times in April 2025 and subsequently described by The Guardian, is still in early development. Neither outlet reported a launch date, and Meta has not publicly acknowledged the effort. The Guardian’s account, published in April 2026, largely builds on the same anonymous sourcing ecosystem as the original FT report rather than citing wholly separate sources. But the ambition is striking: give tens of thousands of workers a simulated version of direct access to the person running the company.
What the reporting actually says
The Financial Times, citing multiple people familiar with the work, described the avatar as a photorealistic 3D representation of Zuckerberg built specifically for internal use. It is meant to converse with staff and provide feedback, functioning as a scaled-up stand-in for the kind of face time most employees at a 70,000-person company will never get.
Importantly, the FT distinguished this avatar from a separate “CEO agent” productivity tool that Meta has also been developing. The two projects appear to serve different purposes: the agent is oriented around workflow tasks, while the avatar is framed as a communication channel. The Guardian’s account aligned with that framing, describing the project’s rationale as helping employees “connect with their boss.” It is worth noting, however, that the distinction between the avatar and the CEO agent tool rests solely on the FT’s characterization; no other source has independently verified where one project ends and the other begins.
No primary documents, internal demos, or on-the-record statements from Meta executives have surfaced. Every subsequent report, from Yahoo Finance to Investing.com, traces back to the same FT sourcing. The volume of coverage reflects media fascination, not a widening pool of independent confirmation.
The technology Meta already has
A photorealistic talking avatar of the CEO sounds futuristic, but Meta has been building toward this for years. Its Reality Labs division has spent heavily on Codec Avatars, a research program that uses machine learning and detailed facial scans to generate 3D digital humans nearly indistinguishable from video. Meta has publicly demonstrated Codec Avatar prototypes in VR settings since 2019, and Zuckerberg himself has appeared in demos showcasing the technology’s fidelity.
On the AI side, Meta launched AI Studio in July 2024, a platform that lets creators build text-based AI characters modeled on real people. AI Studio proves Meta has the conversational-AI infrastructure to power a persona that sounds like a specific individual. Pairing that with Codec Avatar rendering could, in theory, produce exactly the kind of photorealistic, talking digital executive described in the reports.
The gap between a consumer chatbot and a convincing 3D clone of the CEO is still significant. But the building blocks are not hypothetical. They are shipping products and active research programs inside the same company.
Why this is different from a productivity bot
Tech companies have been layering AI into internal operations for years. Coding copilots, meeting summarizers, and automated HR assistants are now routine. What makes the Zuckerberg avatar unusual is its category: it is not a tool that helps employees do their jobs faster. It is a simulation of a personal relationship with the boss.
That distinction matters. A coding assistant does not pretend to be anyone. An AI avatar of the CEO, rendered to look and sound like him, implicitly claims a kind of presence. If an employee asks the avatar about company strategy or layoffs, the answer will carry the weight of Zuckerberg’s likeness even if he never reviewed or approved the response. The line between “AI tool” and “AI spokesperson” gets blurry fast.
Other large tech firms have explored adjacent territory. Nvidia CEO Jensen Huang appeared as a digital avatar during a 2021 keynote, though that was a pre-rendered demo, not an interactive system. Microsoft has invested in lifelike avatar technology through its Mesh platform for Teams meetings. But none have publicly attempted what Meta is reportedly building: a persistent, conversational AI clone of the chief executive deployed across an entire workforce.
The questions employees are likely asking
No reporting to date has included direct reactions from Meta staff. That silence is notable. Meta has cut more than 20,000 jobs since late 2022 and restructured multiple divisions. Employees navigating that environment may not view a digital replica of their CEO as a gesture of closeness. They may view it as a substitute for the real thing.
Privacy is another open thread. If the avatar is conversational, it presumably logs interactions. Who sees those logs? Can an employee ask the avatar a candid question about management decisions without that query being tracked? Meta has not addressed any of these concerns publicly, and the reporting offers no detail on data handling or access controls.
There is also the question of accuracy. Large language models hallucinate. If the Zuckerberg avatar confidently delivers a wrong answer about company policy or project priorities, the consequences could range from confusion to genuine operational harm. The reputational risk of putting the CEO’s face on an unreliable system is not trivial.
Where this stands in spring 2026
As of May 2026, Meta has still not confirmed or denied the project on the record. No public demo has been shown. No internal rollout has been reported. The strongest evidence remains the Financial Times’ original account from April 2025, supplemented by The Guardian’s reporting, which drew on the same broader sourcing ecosystem rather than offering fully independent corroboration. Both accounts are consistent with Meta’s known investments in Codec Avatars and AI Studio.
What is clear is that the technical capability exists, or is close to existing, within Meta’s own labs. What is not clear is whether the company has solved the harder problems: earning employee trust, defining the avatar’s authority, and preventing a communication tool from becoming a source of unease. Building a photorealistic digital Zuckerberg is an engineering challenge. Deciding what it should say, and what it should not, is a leadership one.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
