Top News

Meta is reportedly developing an artificial intelligence version of its CEO, Mark Zuckerberg, that could interact with employees on his behalf, as part of a broader push to embed AI deeper across the company.

The effort is still in early stages, but the idea is straightforward and slightly unsettling, per a Financial Times report. Instead of hearing directly from the founder, employees could eventually engage with a digital version trained to think, speak, and respond like him.

An AI CEO, But Not Quite

The project focuses on building a photorealistic, AI-powered character of Zuckerberg that can hold conversations in real time. This version is being trained on his public statements, internal thinking, tone of speech, and mannerisms. The goal is to make interactions feel close enough to the real person that employees can seek feedback or guidance without needing direct access to him.

Zuckerberg is reportedly involved in training and testing the system himself.

This is separate from another internal idea, often referred to as a “CEO agent,” which is designed more as a productivity tool to assist him with tasks like retrieving information quickly. One is about scaling communication. The other is about augmenting decision-making.

Part of a Larger AI Push

The project sits within Meta’s broader push to restructure itself around AI. The company has been investing heavily to catch up with competitors like OpenAI and Google, with Zuckerberg positioning AI as central to Meta’s future.

Recently, Meta introduced a smaller, purpose-built model designed for tasks such as visual understanding and health-related reasoning, signalling a shift toward more specialised AI systems. Zuckerberg himself has become more hands-on, reportedly spending several hours each week coding and reviewing AI projects.

From Celebrity Bots to Realistic Avatars

This isn’t Meta’s first attempt at AI personalities.

In 2023, the company introduced AI chatbots modelled on celebrities, including figures like Snoop Dogg, allowing users to interact with different personas. It later launched tools that let creators build AI versions of themselves to engage with fans.

But those efforts came with problems. Users began creating inappropriate or unsafe characters, leading to restrictions, especially for younger users. The new approach appears more controlled and more ambitious, focusing on photorealistic avatars that behave closer to real individuals.

The Technical Challenge: Making It Feel Real

Building a believable AI version of a real person isn’t just about training data. It requires real-time voice interaction, facial realism without lag, and contextual understanding. Meta has been investing in voice technology to support this, acquiring companies focused on improving speech and interaction quality. Even then, scaling such systems remains difficult due to the heavy computing power required.

AI Inside Meta, Not Just Outside

The Zuckerberg AI is also part of a wider internal shift. Meta is pushing employees to use AI tools for everyday work, from automating tasks to designing their own agents. Teams are being encouraged to adopt these systems as part of regular workflows.

Some employees, however, see this as more than just a productivity push. There are concerns it could eventually lead to restructuring or job cuts. Meta has said these initiatives are meant to identify skill gaps and improve training, not replace roles.

What This Could Lead To

If the experiment works, the implications go beyond internal use. The same technology could allow creators, executives, or public figures to build AI versions of themselves that can interact with thousands, or millions, of people simultaneously.

In theory, one person could scale their presence infinitely. In practice, it raises a different question: At what point does interacting with a person become indistinguishable from interacting with a system trained to behave like them?

Meta seems determined to find out.