Meta unveils internal 'Zuckerberg' AI

Meta is building an artificial-intelligence version of Mark Zuckerberg that can talk with employees in his place. (ft.com) The project is still in an early stage, and the Financial Times reported on April 13 that Zuckerberg is personally helping train and test it on his speech patterns, tone and current views on company strategy. Quartz, citing the Financial Times, said the system is meant to make workers feel a closer connection to the chief executive. (ft.com; finance.yahoo.com) The Financial Times said Meta recently showed the system to roughly 79,000 employees. The same report said the avatar effort sits inside Meta Superintelligence Labs, the company’s new artificial-intelligence group. (ft.com; finance.yahoo.com) Meta tied that lab to a much larger spending plan on January 28, when it told investors that 2026 capital expenditures would be between $115 billion and $135 billion. In that earnings release, Zuckerberg said he was “looking forward to advancing personal superintelligence” in 2026. (investor.atmeta.com) That spending marks a sharp escalation from Meta’s $72.22 billion in capital expenditures for full-year 2025. The company’s January filing said the increase would be driven by investment supporting Meta Superintelligence Labs and its core business. (investor.atmeta.com) The internal Zuckerberg bot is separate from another automation project around the chief executive’s own work. Quartz, again citing earlier Wall Street Journal reporting, said Meta is also developing a “chief executive officer agent” to handle tasks for Zuckerberg such as pulling up information on demand. (finance.yahoo.com) The company has been moving its artificial-intelligence strategy away from older bets on the metaverse and toward chatbots, models and assistants. Meta said in March that Horizon Worlds would be removed from Quest headsets on June 15, 2026, leaving the social app available only on mobile. (communityforums.atmeta.com; cnbc.com) Meta has been trying versions of “characters” for more than two years. In September 2023, it introduced 28 artificial-intelligence characters, including some tied to celebrities such as Snoop Dogg, Tom Brady and Kendall Jenner, and said creators and businesses would eventually be able to build their own. (about.fb.com) That characters push also drew safety scrutiny. Meta said in January 2026 that it would pause teens’ access to its artificial-intelligence characters across its apps while it built an updated version for younger users. (techcrunch.com; tech.yahoo.com) Meta’s newest public model shows how much the company wants this next phase to look different from its earlier open-source push. On April 8, Meta launched Muse Spark, the first major model from Meta Superintelligence Labs, and CNBC reported that the company plans to keep it proprietary rather than release it openly. (cnbc.com) The result is a company spending like an infrastructure builder while also testing whether a chief executive can be turned into a product for internal use. For Meta employees, the immediate change is simpler: more of Zuckerberg may soon be available, whether he is in the room or not. (ft.com; investor.atmeta.com)