The Financial Times reports that Meta is building an AI version of Mark Zuckerberg - trained on his image, voice, mannerisms, tone, and public statements - that could eventually interact with employees and deliver feedback on his behalf. Meta hasn't confirmed the project, and the FT's account is based on unnamed sources, so details may be incomplete.
This isn't an experiment with consumer-facing AI avatars of the kind D-ID has built for marketing and video creation. This is an internal tool designed to project one person's judgment across an organization of roughly 70,000 employees. The stated goal, per the report, is to give employees access to Zuckerberg's perspective without requiring his actual time.
That framing deserves some scrutiny. An AI trained on public appearances and curated statements will reflect the polished, stage-managed version of a CEO - not the actual person running private meetings or making difficult calls. Employees receiving feedback from this system face a genuine problem: they can't know whether the AI's output reflects real thinking or a statistically averaged impression of it. That gap between representation and reality matters when the feedback affects careers.
The broader pattern here is worth watching. Zuckerberg has a history of using himself as a product demo - most notably the AI-generated version of himself Meta used to introduce Meta AI in 2023. Building an internal clone fits that same instinct. But using AI avatars for marketing is different from using them as a management layer.
If the project moves forward as described, it sets a precedent other large organizations will notice. The harder questions - who is accountable for feedback delivered by an AI proxy, and how employees are supposed to respond to criticism they can't verify came from a human decision - don't have obvious answers yet.