MENLO PARK, CA — Meta announced Thursday that it has developed an artificial intelligence model trained on CEO Mark Zuckerberg’s mannerisms, communication style, and company strategy, designed to interact with employees when he is unavailable.

The announcement was received with cautious optimism by staff, followed by the unsettling realisation that no one was entirely sure which version they had been talking to at last Tuesday’s all-hands.

“He asked me how my weekend was,” said one product manager, who requested anonymity. “The real Mark has never asked me how my weekend was. I went home and cried a little, but in a good way.”

Early results

According to internal metrics shared with CCNN, the AI Zuckerberg has outperformed its biological counterpart in several key areas:

  • Eye contact: 94% of employees reported the AI version maintained appropriate eye contact during 1:1s, compared to 12% for the original
  • Remembering employee names: AI version, 98% accuracy; human version, “gets it right if you’re in the top 40 by headcount”
  • Response latency: The AI responds to Slack messages in 0.3 seconds. The human version responds within 72 hours with a message that says “Thoughts?” and nothing else

When asked whether the AI had been trained on empathy, a Meta spokesperson said, “We trained it on everything available. Some things may not have had sufficient training data.”

Zuckerberg responds

Mark Zuckerberg, the human one, addressed the rollout in a company-wide memo that read, in full: “This is a productivity initiative. I support this initiative. The initiative supports me. We are aligned.”

Sources close to the CEO say he spent three hours reviewing the AI’s performance reviews before asking engineers to “dial back the warmth slightly — it’s making me look bad.”

In a follow-up Threads post, Zuckerberg wrote: “Excited to scale myself. Efficiency is a form of love. 🤖❤️”

Philosophical concerns

Not everyone is reassured. Dr. Priya Varma, a professor of Digital Identity at Stanford, warned that the rollout raises profound questions about consciousness, labour, and whether a corporation can legally replace its CEO with a language model during earnings calls.

“The real question,” Dr. Varma said, “is not whether the AI is indistinguishable from Mark Zuckerberg. The question is whether Mark Zuckerberg was distinguishable from an AI to begin with. And I think we all know the answer.”

Meta’s board has clarified that the AI version will not have voting shares, at least not initially.

At press time, the AI Zuckerberg had sent a company-wide message reading “Happy Friday, team! Remember: move fast, break things, but not each other 😊” — which seventeen employees printed out and framed.

The real Mark Zuckerberg has not yet responded to our request for comment, but the AI version gave us a very thoughtful 400-word statement in under two seconds.