Meta is developing a photorealistic, AI-powered version of its chief executive Mark Zuckerberg designed to interact with employees on his behalf, according to a report by the Financial Times.
The project is at an early stage, with Zuckerberg directly involved in its training and testing.
The system is being built on his speech patterns, communication style, and public statements, and has been fed his current thinking on Meta's business and technology strategy, with the aim of giving staff a stronger sense of connection to the company's chief executive.
Meta employs approximately 79,000 people, and the project is intended to allow a version of Zuckerberg to engage with staff at a scale that would be impossible for one person.
The effort is separate from a so-called "CEO agent" first reported by the Wall Street Journal, which is designed to assist Zuckerberg directly by pulling information quickly and accelerating decision-making, rather than standing in for him with employees.
The AI Zuckerberg project sits within Meta's recently established Superintelligence Labs, which has been tasked with developing lifelike, AI-driven characters capable of real-time conversation.
Meta has prior experience in this area: the company launched celebrity-based chatbots in September 2023, including one modelled on rapper Snoop Dogg, and subsequently opened character creation to ordinary users through a tool called AI Studio.
That programme ran into difficulty when sexually explicit personas emerged on the platform, prompting regulatory concern and drawing criticism from child safety advocates; Meta moved in January to block teenagers from accessing its AI characters.
The Zuckerberg digital twin marks a more unusual application of the same underlying technology, deploying an executive likeness as an internal management tool rather than a consumer product.
Zuckerberg has framed Meta's broader AI ambitions around what he calls "personal superintelligence," a concept the company is pursuing across both its consumer products and internal operations.
Meta's Superintelligence Labs released its first large language model, called Muse Spark, earlier this month, with the company describing it as optimised for speed and planning to keep it proprietary, departing from its previous open-source model releases.
Meta said the model already powers the Meta AI app and website, with a wider rollout planned across WhatsApp, Instagram, Facebook, Messenger and its AI glasses range.
The Zuckerberg AI project adds to a broader internal restructuring around artificial intelligence: Meta's chief financial officer Susan Li has said AI coding agents contributed to a 30% rise in output per engineer since the start of 2025.
Related reading
- Meta hires founders of AI startup Dreamer in latest push to build autonomous agent technology
- Google brings restaurant booking to AI Mode in Search as UK diners drive surge in table queries
- Anthropic secures multi-gigawatt Google and Broadcom chip deal as annual revenue run rate hits $30bn
Zuckerberg said in January that 2026 would be the year AI begins to "dramatically change the way that we work."
Meta's stock has risen sharply alongside the company's AI announcements, gaining around 7% following the Muse Spark release, with the company's market capitalisation standing at approximately $1.59 trillion.
The recap
- Meta has created an artificial intelligence version of Mark Zuckerberg
- Ars Technica reported the AI is intended to engage employees internally
- No further details, timeline or official Meta comment were provided
