Meta Chief Executive Mark Zuckerberg published a Mark Zuckerberg AI essay in The Wall Street Journal’s opinion section arguing that future superintelligence should be spread widely rather than controlled by a small set of institutions.
Zuckerberg wrote that he expects people within the next few years to use AI systems more capable than humans to create products, learn, start companies and improve parts of daily life. His central claim was political as much as technical: the main fight, in his view, is over who gets to use those systems.
The essay, published July 28, 2026, set out three principles Zuckerberg said should guide Meta’s work on superintelligence: individual empowerment, invention and a balance of power. He said Meta is committed to building around those ideas.
What did Mark Zuckerberg say about AI?
Zuckerberg argued that superintelligence should be treated as a personal tool available to many people, rather than as a capability reserved for governments, labs or a few large organizations. In his framing, broad access would let people pursue their own goals and would also keep powerful institutions from becoming the only actors able to use advanced AI.
He criticized AI developers and commentators who focus heavily on catastrophic scenarios. Zuckerberg said he does not understand why someone who believes AI will erase most work and human relevance would still race to build it. He also warned that treating extreme centralization as the only safe answer creates its own danger.
To make the argument, Zuckerberg reached for a political analogy rather than a benchmark chart. He wrote that democratic and economic history show people do not agree on one universal definition of a good life, so a single expert group should not decide what advanced AI should optimize for everyone.
How Zuckerberg connects AI access to safety
Zuckerberg’s safety argument is that power can check power. If a small number of institutions hold superintelligence, he wrote, they could shape economics, science and politics even if they start with good intentions.
He used a legal example: one person with a superintelligent lawyer would gain an unfair edge in court, while universal access to similar help could make legal outcomes more efficient and fairer. The example is a thought experiment, not a claim that such tools already exist.
Zuckerberg also rejected the idea that one benevolent superintelligence could be aligned with all of humanity’s interests at once. His argument is that human values conflict, so any single system or authority would have to favor some priorities over others.
What risks did Zuckerberg acknowledge?
Zuckerberg did not argue that advanced AI carries no risks. He wrote that cybersecurity and biological misuse need serious attention, while drawing different lessons for each area.
For cybersecurity, he said the history of open-source software suggests broader access to powerful systems can improve safety and security over time. For biological risks, he said governments and other institutions should coordinate more on responsible deployment of capable models.
On jobs, Zuckerberg said the economic outcome depends on whether AI is used mainly to automate work or to help people increase their skills and create businesses. He warned that a tilt toward automation could hurt employment and the economy. If superintelligence is widely distributed, he said he expects more jobs, easier company formation without large amounts of capital, and more people working in small businesses.
The essay is a statement of Meta’s preferred direction, not evidence that personal superintelligence has arrived. Zuckerberg is staking out the company’s argument before the technology, the regulation and the damage reports are all in.
This story draws on original reporting from The Wall Street Journal.
