Menlo Park, California, April 14, 2026, 06:07 PDT
- Meta is working on an AI-powered Mark Zuckerberg, according to an April 13 report in the Financial Times. The virtual version is being trained with his likeness, voice, gestures, and public remarks, and will be used for employee interactions.
- Meta wrapped up 2025 reporting a headcount of 78,865. For 2026, the company is projecting capital expenditures between $115 billion and $135 billion, pointing to AI infrastructure and talent as the main drivers for the increase.
- Meta has started moving leading engineers over to its new Applied AI group, according to an April 9 report from Reuters, as Zuckerberg champions “AI-native tooling” and a leaner team structure. Reuters
According to the Financial Times, Meta is working on an AI-powered replica of CEO Mark Zuckerberg that would handle conversations with staff on his behalf. The company, which owns Facebook and Instagram, is pushing further into AI by training the tool on Zuckerberg’s likeness, voice, style, and public remarks. Employees could eventually interact with this digital proxy rather than the founder himself.
The shift is significant: Meta’s AI efforts are increasingly focused internally, not just on products for users. Last week, Reuters said Meta began reassigning leading engineers to a fresh Applied AI group. Back in January, Zuckerberg told investors the company was spending on “AI-native tooling” and “flattening teams.” Reuters
The numbers stand out. Back in January, Meta reported a headcount of 78,865 as of the end of 2025, and mapped out $115 billion to $135 billion in capital expenditures for 2026—most of that earmarked for infrastructure and AI-focused technical hires. Zuckerberg, for his part, talked up plans to push “personal superintelligence” forward in 2026. Meta Investor
That’s yet another indicator of Meta’s shift away from its metaverse ambitions. According to Engadget, the company had already spent a while developing photorealistic, 3D-animated AI characters. The Guardian noted that lately, those resources have been steered toward making a Zuckerberg avatar instead.
Meta started its public push earlier this year. The launch of AI Studio in 2024 gave Instagram creators tools to make AIs that could handle routine DMs and reply to stories “as an extension of themselves.” The company promised those bot replies would get clear labels. About Facebook
According to The Verge’s readout of the FT report, Zuckerberg has been hands-on with the avatar project, reportedly logging five to 10 hours each week on coding and technical reviews for Meta’s AI initiatives. This work is distinct from the personal “CEO agent” the Wall Street Journal said in March he’s developing to support his day-to-day as chief executive. The Verge
The AI competition keeps heating up. Meta just pushed out its Muse Spark model—its first new AI in roughly a year, Reuters said on April 8. Tests put Muse Spark on par with Google, OpenAI, and Anthropic in several respects, though it’s still weaker when it comes to coding skills and abstract reasoning. Alexandr Wang, head of Meta’s superintelligence division—industry shorthand for those aiming to build AIs that eclipse human abilities on most tasks—admitted there are “certainly rough edges” that need attention. Reuters
A digital boss avatar brings new complications. Back in January, Meta pulled the plug on teen access to its AI characters worldwide, aiming to launch a version with parental controls; lawsuits alleging harm to young users are ongoing. The company maintains that AI replies from creators must be clearly marked. If an internal Zuckerberg bot appears, expect even tougher scrutiny—issues around transparency, trust, and liability get sharper when software stands in for the CEO.
Within Meta, there’s no ambiguity on course. “AAI is one of the company’s highest priorities,” Maher Saba, the executive leading the newly formed AI engineering team, wrote in a memo reviewed by Reuters. Staff moves into the unit, Saba made clear, aren’t up for debate. From its avatar-driven past, Meta now faces a different hurdle: whether employees will accept an artificial intelligence system as a boss. Reuters
