Zuckerberg Calls Concentrated AI Power the Real Danger

Mark Zuckerberg picked a fight in the op-ed pages this week, and the target is the way his most-hyped rivals have been talking about safety. In a Wall Street Journal piece on July 28 titled "The AI Future Is for Everyone," covered by The New York Times, the Meta chief argued that the real danger of superintelligence is not that it exists but that access to it will be locked inside a handful of institutions. His line, quoted across the coverage, was that "the notion that AI is so dangerous that the only safe path is an extreme concentration of power seems dangerous."

The names he did not quite name are Anthropic and OpenAI. Zuckerberg said building AI "in a tightly controlled manner, as Anthropic and OpenAI have done" would amount to "abandoning our values," a jab, per The Hill, at Anthropic's Dario Amodei and OpenAI's Sam Altman, both of whom have spent the last year publicly worrying about catastrophic risk. Meta's counter-pitch has a slogan, "personal superintelligence," organised, Zuckerberg wrote, around three principles: individual empowerment as the source of prosperity, invention as the primary purpose of superintelligence, and balance of power as the foundation of safety.

The reason this matters beyond the usual CEO sparring is that it is also a policy play. The op-ed landed the same day, Forbes reported, that more than a thousand employees at leading AI labs including OpenAI and Anthropic signed an open letter asking the U.S. government to build tools that could deliberately slow AI development. A competing letter signed by Meta, Nvidia, Microsoft and Palantir has asked regulators not to restrict open-weight model formats. The philosophical case for distributed intelligence is also the commercial case for Meta's release strategy.

The honest caveat is that Meta is centralising its own AI work at the same time it warns against centralisation elsewhere. Zuckerberg has just consolidated Meta's model, product and research teams under a new Meta Superintelligence Labs, with former Scale AI CEO Alexandr Wang installed as chief AI officer. What the reporting does not give you is any concrete policy ask, any specific rule Zuckerberg wants Washington to reject, or a plan for how personal superintelligence reaches consumers at scale. Take the framing as a positioning move, not a manifesto with attached legislation.

If the argument sticks, the winners are developers, startups and enterprises that would rather run open-weight models themselves than rent frontier capability from three vendors. Watch whether Meta keeps releasing weights as its own models climb into genuinely frontier territory. That is where the "for everyone" line stops being rhetoric and starts costing something.