Zuckerberg says superintelligence should belong to everyone, not a few companies
New Delhi: Meta CEO Mark Zuckerberg has laid out a broad vision for superintelligence that puts personal AI agents, open access and individual control at the centre of the company’s strategy. In a lengthy note published on August 10, Zuckerberg argued that advanced AI should be widely distributed instead of being controlled by a small number of companies, governments or institutions.
He said Meta plans to build personal superintelligence systems that can help with work, health, learning, finances, relationships and daily tasks. The company also plans to offer free or affordable access to some of these systems and create a private mode where, according to Zuckerberg, even Meta cannot see or grant access to a user’s personal information.
Today we’re also opening the weights for Muse Glimmer, a great 30B parameter dense model that can run locally. Soon we’ll also release the weights for Muse Spark 1.2, our latest foundation model. Meta is a strong supporter of open source and I’m proud of these releases. Congrats…
— Mark Zuckerberg (@finkd) August 10, 2026
Zuckerberg says AI should focus on invention
Zuckerberg argues that the main value of superintelligence will come from helping people create and discover new things.
“Invention, not automation, will be the greatest contribution of superintelligence,” he wrote.
He expects AI agents to act as personal assistants that work continuously, learn a person’s goals and help manage tasks across different areas of life.
Zuckerberg also said these systems could change education by giving people access to personalised tutors, and could speed up scientific research in areas such as drug discovery and biology.
Meta expects more jobs, smaller companies
The Meta chief pushed back against predictions that advanced AI will simply remove large numbers of jobs.
He expects AI to help individuals build businesses with smaller teams and create new types of work. Zuckerberg wrote that future companies could become smaller in size but more numerous, with personal AI agents allowing a few people to operate businesses at much larger scale.
He also said Meta will invest in training programmes for construction and technical workers as data centre and energy infrastructure expands.
Open source remains part of Meta’s AI strategy
Zuckerberg said Meta remains committed to open source AI and plans to resume releasing some open models through Meta Superintelligence Labs.
He also criticised policies that could slow US AI development, saying American companies need to move quickly to compete with China and other countries.
At the same time, he called for closer cooperation between AI labs and governments. One proposal would see frontier AI companies share intermediate model checkpoints with governments before final releases so security teams can test new capabilities and strengthen critical systems.
Meta plans new oversight for AI releases
Zuckerberg said Meta is changing how decisions around powerful AI models are made.
The company’s independent board of directors will have authority to approve safety criteria for model releases and review whether individual releases meet those standards.
He also said no single person, including himself, should control decisions around superintelligence deployment.
Zuckerberg’s wider argument is that AI safety should rely on distributing capability across people and institutions rather than concentrating it in one place. Meta is now placing that idea at the centre of its next phase of AI development.