Meta builds internal 'Zuckerberg' AI
The Financial Times reports Meta is developing an internal AI version of Mark Zuckerberg to interact with staff as part of a broader push toward ‘personal superintelligence’. Meta’s chief AI scientist Yann LeCun has publicly downplayed recent vulnerability claims about Anthropic’s Claude, calling the controversy overblown. (ft.com) (storyboard18.com)
Meta is building an internal artificial intelligence version of Mark Zuckerberg to talk with employees on his behalf, according to the Financial Times. (ft.com) The report said Meta has been developing photorealistic three-dimensional AI characters that can speak in real time, and Zuckerberg has been training and testing his own “character” for internal use. The system is meant to reflect his voice, mannerisms, and recent views on company strategy. (ft.com) The project lands five days after Meta introduced Muse Spark on April 8, its first model from Meta Superintelligence Labs, the unit leading Zuckerberg’s push to rebuild the company’s AI effort. Meta said Muse Spark will power Meta AI across WhatsApp, Instagram, Facebook, Messenger, and its AI glasses. (about.fb.com) Meta has framed that effort around “personal superintelligence,” Zuckerberg’s term for AI assistants that know users well enough to help with goals, decisions, and daily tasks. The company says that vision will sit at the center of its consumer products rather than only enterprise software. (meta.com) An internal Zuckerberg bot would extend that logic inside Meta itself: instead of only answering consumers’ prompts, the company is testing whether an AI stand-in can answer employee questions and relay leadership guidance at scale. The Financial Times said the tool is being built for staff interaction, not public release. (ft.com) The timing also overlaps with a separate fight over how dangerous advanced AI systems already are. On April 7, Anthropic launched Project Glasswing and said its Claude Mythos Preview could identify and exploit zero-day software vulnerabilities across major operating systems and browsers, so access would be limited to selected partners. (anthropic.com) Meta’s chief AI scientist Yann LeCun rejected much of that alarm. Storyboard18 reported that LeCun described the reaction to Claude Mythos as exaggerated “drama,” arguing that the claims were being overstated. (storyboard18.com) Anthropic has taken the opposite line in public. In its own technical write-up, the company said Mythos Preview could identify and then exploit zero-day vulnerabilities in every major operating system and every major web browser when directed by a user, which is why it is not being released broadly. (red.anthropic.com) Those two threads — building a more personalized AI inside Meta and dismissing some rival safety warnings outside it — show how Zuckerberg’s company is trying to move faster on AI products while resisting arguments for tighter limits on open models. The next test is whether Meta keeps these tools as internal assistants, or turns them into a template for how executives themselves get replicated in software. (ft.com)
