Zuckerberg's 6,500-word AI utopia keeps reminding us exactly how it goes wrong

Mark Zuckerberg’s 6,500-word manifesto on personal AI superintelligence isn’t an anti-doomer screed so much as a detailed list of reasons to be nervous. The Meta CEO paints a future where abundant intelligence checks and balances itself into positive outcomes, but his examples keep colliding with uncomfortable realities.

The essay arrives at a brutal moment for tech credibility. A recent survey shows 64% of Americans believe social media has harmed democracy, and just this weekend Meta was fined $567 million in a case about harm to children. The public does not trust tech executives to shepherd powerful new technologies responsibly. Instead of acknowledging that erosion of trust, Zuckerberg’s philosophical generalities about intelligence sound eerily similar to the hazy free-speech abstractions he and Jack Dorsey once deployed to hand-wave away social media’s downsides.

Take his vision for education. Zuckerberg describes personalized AI tutors with PhD-level knowledge and unlimited patience, democratizing learning for everyone. But the product he’s describing already exists in ChatGPT, Claude, and Gemini. The dominant use case in schools right now isn’t learning—it’s cheating. Students use these tools to complete homework and write essays, and without robust watermarking, teachers can’t distinguish AI-generated work from authentic student effort. It’s a real, present harm that the essay simply ignores.

The legal example is even stranger. Zuckerberg argues that if only one person had a superintelligent lawyer, it would be unfair, but if everyone had one, the system would self-correct. That logic sidesteps the question of who trains these systems, what biases they encode, and whether equal access to a technology that can generate plausible-sounding legal arguments actually produces justice—or just a higher volume of confident nonsense filed in court dockets.

This is the pattern that makes the essay unsettling. Zuckerberg frames every concern as a distribution problem that more AI will solve, without grappling with the ways AI is already being misused. The disconnect between the utopian framing and the on-the-ground reality isn’t just a PR problem for Meta. It’s a case study in why the public remains skeptical that the people who built social media should be the ones deciding what superintelligence looks like.