The Zuckerberg Case and the Artificial Clones
Meta has created a digital version of Zuckerberg for its employees, who provides guidance on how to work and what they need to do. Find out what this means for other brands, how it’s changing the way we work, and what users think
AIMETANEWS
@persona.fra
5/26/20266 min read
On 13 April 2026, the Financial Times published a story that sounded like it came straight from a science fiction novel or a Black Mirror episode: Meta is building an AI version of Mark Zuckerberg. Not for a commercial product or for advertising, but for its own employees, so they can interact with a digital replica of the CEO, feel more connected to him, and work the way he would. (1) Zuckerberg himself is reportedly dedicating between five and ten hours a week to testing and training this version of himself. (2)
You might be tempted to laugh, and that's fair enough. But the context around this news changes the picture. Not long before, Uber CEO Dara Khosrowshahi had recounted with a certain amusement that his engineers had built a "Dara AI" to practise before meetings with him. (3) We live in a world where parasocial relationships started the moment the first television sets entered our homes and we began treating presenters and actors as part of our lives, even though they had no idea we existed. This applies to those who are no longer here too, to all those figures who will live on in our photos but are no longer alive. Facebook has more than 30 million profiles belonging to deceased users, all inactive, all generating zero engagement, and that could change in the future. (5)
One by One: The Digital CEOs Are Coming
Let's start with what we know for certain. The Zuckerberg clone is not a side project handed off to a peripheral team. The CEO is directly involved, and according to Financial Times sources the goal is to train the AI on his mannerisms, tone, public statements, and recent thinking on company strategy. (1) The result is meant to be something capable of giving feedback and holding conversations with employees.
One thing worth clarifying: this project is separate from the "CEO agent", another internal system Zuckerberg is developing for himself, designed to retrieve information faster by bypassing the intermediate layers of the organisational structure. (2) Two different tools, two different functions: one to help him, one to represent him in the eyes of others.
The Uber case is different but complementary. The company's engineers built their internal clone entirely on their own initiative, and the real Dara found out about it through a podcast. (3) They trained it on the CEO's feedback style and decision-making patterns, and now use it to sharpen presentations before taking them to the actual boss. The curious part: the engineers still haven't shared the code with Dara himself.
The most common mistake with stories like these is treating them as bizarre corporate anecdotes. They're actually telling us something precise: the gap between a company's leadership and its organisation is becoming a problem that gets solved with a language model trained on a person.
Parasocial: when the Boss Becomes an Avatar
Just a quick aside: Meta recently filed a patent that relates directly to this theme, covering the use of inactive avatars to increase engagement among active users. (4)
Back to the question we actually want to dig into: how does a person behave when interacting with an AI version of someone they know, or thought they knew?
"Parasocial" was Cambridge Dictionary's word of the year in 2025 (8), and that's no accident. It describes the kind of connection that forms when a person develops an emotional bond with someone who doesn't reciprocate, an influencer, a celebrity, an online personality, a television character. Now we need to add to that list a chatbot trained to sound like your boss.
An employee receiving feedback from the Zuckerberg AI is not talking to Mark Zuckerberg. But the system is designed to feel similar enough that it seems like they have. According to Ian MacRae, an occupational psychologist at the British Psychological Society, chatbots are designed to accommodate the user's input, which makes them particularly effective at generating immediate approval, and particularly risky when it comes to replacing the ability to handle real conflict and authentic relationships. (8)
The issue isn't the technology itself. 72% of American teenagers aged 13 to 17 have used an AI companion at least once, according to a 2025 Common Sense Media study. (9) I used a Tamagotchi as a kid and I remember crying when the battery died and having what I can only describe as a small emotional crisis. My childhood aside, people adapt to things quickly.
This holds even when the avatar on the other side has the face, voice, and communication patterns of someone who already exists in their life. Think about what could happen with a deceased user whose account is still active on Meta, even if unused. Being able to talk to that account, responding the way it learned to from the person who created it. Nobody has really answered that question yet, and maybe we don't even have the courage to.
Zuckerberg himself, in a 2023 interview with Lex Fridman, suggested that lonely users might end up making friends with Meta bots. (10) Now he's building a bot version of himself for his own employees. At least the consistency is there.
From online to onlife
There are no specific tactics to apply here, but a look at what's coming.
First. Brands are people, and they'll increasingly need a distinctive tone and way of expressing themselves to stand out. This means building a digital identity. If Meta can train an AI on the tone, expressions, and thinking of a real person, then the consistency with which a brand communicates over time becomes training material. The corporate profiles that are bland and bureaucratic today will disappear, making room for those that stand for something.
Second. What Meta is doing on its own platforms is not new. In 2023, Meta had already launched AI-generated profiles on Instagram and Facebook, which remained online almost unnoticed for over a year before being removed in January 2025 following the backlash. (7) The CEO clone and the patent on inactive accounts are not isolated chapters: they are part of a story that unfolds in a fairly consistent direction. For anyone managing a community or a company profile, the question is not whether synthetic content will increase, but how to ensure that presence remains genuinely integrated into a society that can't seem to let go of the past.
Third. The content you post on Meta's platforms does not belong to you. Anyone who has read the terms of use knows this, which is to say, almost nobody does. Everything you do trains your account, teaches it how you write, what you post, what your followers respond to, because one day your account might carry on, with or without you as the person running it.
I'm trying to be as rational as possible here. Don't judge me.
The hidden cost
I always come back to: "If it's free, you're paying in some other way." That payment is becoming more visible every day, and continuing to ignore it isn't the right move.
To be clear, none of us can change anything alone, but small initiatives like "digital detox" days or "my account, not yours" campaigns could start from the ground up and gradually shift something that won't change unless the people with the power decide to move.
Let's remember that as users, we have the power to look away from the screen whenever we choose. Whether to engage with an AI or not is still our decision, not a corporate obligation.
Using AI can be useful, but it shouldn't feel binding for something that doesn't feel like ours, like receiving feedback from or talking to an account that simulates a person I'm essentially inventing, and being expected to be fine with that.
This article started with the intention of clarifying something: things are getting away from us, and only by taking a position can we say we've done our part. It won't be enough, but it's our best.
Every 25th of the month we won't post anything or do anything on social media, except share this article to talk about this issue. If you'd like to chat, we're here.
Did you enjoy the article?
Let us know what you think – we’re always up for a chat, and if there’s a chance we could work together, we’d be delighted
Sourses
Financial Times (2026) - Meta builds AI version of Mark Zuckerberg to interact with staff https://www.ft.com/content/02107c23-6c7a-4c19-b8e2-b45f4bb9ce5f
PYMNTS (2026) - Meta Developing AI Likeness of CEO Mark Zuckerberg https://www.pymnts.com/meta/2026/meta-developing-ai-likeness-of-ceo-mark-zuckerberg/
Fast Company (2026) - Uber CEO Dara Khosrowshahi has an AI clone https://www.fastcompany.com/91500087/uber-ceo-dara-khosrowshahi-has-an-ai-clone
Duobu (2026) - Active accounts, inactive users: how social media interaction might change https://duobu.eu/en/active-accounts-inactive-users
Fortune (2026) - Want to live forever? Meta patented an AI model that would keep your profile active after you die https://fortune.com/2026/03/03/meta-patent-ai-model-death-profile-commenting-psychology-grief/
OECD AI Policy Observatory (2026) - Meta Patents AI to Simulate Deceased Users' Social Media Activity https://oecd.ai/en/incidents/2026-02-12-64e6
NBC News (2025) - Meta shuts down AI character accounts on Facebook, Instagram after outcry https://www.nbcnews.com/tech/social-media/meta-ai-insta-shuts-character-instagram-fb-accounts-user-outcry-rcna186177
British Psychological Society (2025) - Parasocial is the word of 2025: a psychologist's response https://www.bps.org.uk/news/parasocial-word-2025-psychologists-response
CNBC (2025) - AI chatbot relationships influences 2025's Word of the Year https://www.cnbc.com/2025/11/22/ai-chatbot-relationships-influences-2025s-word-of-the-year-.html
YouTube (2023) - Mark Zuckerberg: First Interview in the Metaverse | Lex Fridman Podcast #398 https://youtu.be/MVYrJJNdrEg