Meta chief executive Mark Zuckerberg is betting that superintelligence should not be controlled by a handful of companies, governments or experts, arguing that advanced artificial intelligence (AI) should instead be placed directly in the hands of billions of people.

In a lengthy essay published 10 August, Zuckerberg laid out Meta’s philosophy for what he described as an approaching era of AI exceeding human capabilities, built around three principles: individual empowerment, invention and a balance of power favoring people.

“The defining questions of our age are who will have access to superintelligence and what will we direct it towards,” Zuckerberg wrote, questioning whether it would be centralized among a few institutions or broadly available.

The Meta founder also pushed back against the increasingly bleak debate surrounding AI, particularly predictions that it could eliminate most jobs or diminish humanity’s relevance.

Instead, Zuckerberg argued that AI’s greatest contribution would be invention rather than automation, from discovering new medicines to helping people build businesses and create new products.

AI agent for everyone

Central to Meta’s vision is a personal AI agent that understands its user, goals and interests and operates continuously on their behalf.

Zuckerberg said such agents could help manage careers, finances, health, relationships and everyday tasks. Meta also envisions personalized tutors, creative tools and AI systems capable of helping individuals launch businesses with fewer employees and less capital.

Meta intends to offer free versions accessible to billions, with users paying for additional computing power if needed.

Zuckerberg predicted this could create more businesses and eventually more jobs, rather than the mass unemployment feared by some AI critics.

He acknowledged, however, that disruption is inevitable.

Jobs will change and companies could become smaller as AI allows fewer people to accomplish more. But Zuckerberg argued that history suggests new technologies also create entirely new occupations and industries.

AI risks remain

The optimistic pitch came with a lengthy list of dangers.

Zuckerberg acknowledged risks involving job displacement, cybersecurity, biological and chemical threats, government surveillance and the possibility that humanity could eventually lose control of superintelligent systems.

On biological threats, he acknowledged that technology capable of helping scientists design medicines could also help malicious actors create harmful compounds. He argued governments should focus on controlling the physical production and distribution of dangerous materials while accelerating defenses and medical treatments.

Meta also wants frontier AI companies to share intermediate versions of advanced models with the US government before training is complete, giving authorities time to identify vulnerabilities and strengthen critical infrastructure.

Don’t centralize AI

Zuckerberg’s broader argument is that concentrating superintelligence could itself become a danger.

He compared it to giving only one person a superintelligent lawyer or cybersecurity system: whoever controls it would gain overwhelming power. If everyone has comparable tools, he argued, those competing interests could act as checks and balances.

Meta said it would continue supporting open-source AI models while building privacy protections for personal agents.

Zuckerberg also said he should not have sole authority over how Meta releases superintelligence. The company is giving its independent board power to approve safety criteria and review whether future model releases meet them.

For Zuckerberg, the objective is ultimately simple but enormous: Make superintelligence powerful enough to transform civilization without allowing that power to become concentrated enough to control it.

“The future will be created by all of us,” he wrote.