Mark Zuckerberg told the Sources podcast that Meta‘s personal AI agent, Muse, will be free for a large amount of use, and that the way to a good future with AI is to spread the technology widely rather than keep it in a few labs. Everything below is what he said in the interview, and much of it is Meta’s own plan and Zuckerberg’s own opinion, not independent evidence.
Why he argues for distributing AI
“My view on this is that the path to have a positive future for everyone is to make sure that we distribute the technology as widely as possible,” Zuckerberg said. He set out three principles: that empowering people is the source of prosperity, that AI’s primary purpose will be the invention of new things, not automation, and that safety comes from establishing the right checks and balances and balance of power rather than restricting access.
He said he is more worried about a small number of labs or people having control than about broad access. On security he argued that “the best antidote to someone having an AI that could potentially hack into systems is having everyone have access to an AI so they can harden their own systems first”, pointing to the history of open source software, where scrutiny by many people tends to produce a more stable environment. He also said he has “an allergy” to the idea that a small group of experts should decide what AI does for big problems, because people know best what matters in their own lives.
What Muse does
Zuckerberg described Muse as a proactive agent that works on projects and suggests new ones. His own examples were personal. He asked it to set up a weekend baking project suitable for a three-year-old and an adult who knows nothing about baking, and to use Instacart to get the ingredients. He has put cameras in his MMA gym and asks it to watch them and send feedback. He also had it build a strategy guide for the computer game Civilization, which it later expanded into an app that teaches historical lessons.
Pricing, and what is still unclear
On money, Zuckerberg said Muse will be able to “make people money and save people money”. Meta plans to offer, to start, about 100 million tokens a week for free along with a virtual machine, and to take a very small cut of the transactions the agent enables, paid by the businesses involved rather than the person buying. He said Meta is working with Stripe on payments.
On privacy, he said Meta can commit that it cannot see the content in a user’s Muse, and that this can be verified technically through a confidential virtual machine. He said more would be published “in the coming weeks” as the product rolls out more widely. That means the privacy claim is a promise that has not yet been independently tested, and he did not say when Muse will reach South Africa.
Data centres and skills
Asked about Meta’s data centres, Zuckerberg said they can be good for local communities through tax revenue and jobs, citing a Louisiana example where tax revenue funded $50,000 bonuses for teachers. He also described Meta’s America’s Workforce Academy, which aims to train more fibre technicians, electricians and skilled carpenters because there are not enough people to build the centres.
What this means for South African owners
Nothing here is available to a South African business today, so treat it as a preview. If Muse arrives, the practical questions are the ones any owner should ask of an agent that handles money: what it can access, what it costs after the free allowance, and who is liable if it makes a mistake. Because the commerce cut is paid by businesses, compare it with what Meta already charges elsewhere, such as its WhatsApp service messages, and check the pricing before you build a process on top of it.
For more on how Meta describes its products, see Meta’s own site, and for background on the man himself, Mark Zuckerberg on Wikipedia. More on AI for business is in our Tech & Telco section.
