Editorial Team
7 min read
Mark Zuckerberg says Meta will pursue personal superintelligence for billions of people, with free versions, paid compute and private agents. The plan turns on capacity, release rules and governance that the company has not yet specified.
Mark Zuckerberg, Meta’s founder, chairman and chief executive, has made a case for “personal superintelligence” as an alternative to leaving the most capable AI inside a small group of companies, governments and other institutions. In his manifesto, Zuckerberg says Meta will build agents aligned with individual users’ goals and make free versions available to billions of people.
It is a strategic declaration, not a product launch. Meta does not say that it has deployed a superintelligent agent, nor does the statement define the capabilities of the free service, a launch date or the terms of its promised private mode. The proposal matters because its answer to AI concentration is not merely to release technology: it is to build a consumer service whose capacity, privacy settings and release decisions will still be made by Meta.
Meta’s official Muse Glimmer announcement describes an open agentic model designed to run on a device. Source: research.meta.ai.
Zuckerberg’s argument starts with access. He says the consequential question is not whether superintelligence will exist but who can use it, and he contrasts personal agents with AI aimed at companies, governments and institutions. The earlier published essay sets out the same three principles: individual empowerment, invention rather than automation, and a balance of power.
The proposed agents would assist with tasks ranging from learning and health to business creation. Zuckerberg predicts that broad availability could make it easier to start companies without raising substantial capital and lead to more jobs over time. He also says the outcome could be negative if AI’s use shifts toward automation. Those are competing forecasts about how systems will be deployed, not observed employment results.
The manifesto makes a concrete distribution promise with a built-in constraint: Meta says it will offer free versions, while people who want more compute would use a “dynamic auction mechanism.” It says the aim is the lowest possible price for the intelligence and compute being used while capacity goes to what people collectively value. That describes a market for finite capacity, rather than equal or unlimited access.
The company says it will build a fully private mode in which Meta or another provider cannot see or grant access to a user’s information. That is a product commitment; the announcement does not describe its technical design, scope or audit method.
Meta also says it will resume releasing some open-source models now that Meta Superintelligence Labs is operating. A report on the announcement says Meta plans to open the weights of Muse Glimmer and a version of Muse Spark 1.2. Open model weights can make a model available beyond Meta’s own service, but they do not establish the availability, price or privacy of the personal agents promised in the manifesto.
The distinction has policy significance. Meta and Microsoft had already joined a July letter urging U.S. policy support for open-weight models, while OpenAI and Anthropic did not initially sign, according to contemporaneous coverage. The manifesto’s claim that distribution is safer is therefore also an intervention in a live argument over how advanced models should be released and governed.
Zuckerberg describes Meta as founder-controlled while arguing that no one person should decide how superintelligence is deployed. His proposed safeguard is a governance structure under which Meta’s independent directors approve safety criteria for model releases and review whether releases meet them.
The authority structure matters. Meta’s 2025 annual report says Zuckerberg can exercise voting rights over a majority of the company’s outstanding voting power, giving him control over stockholder votes, including director elections and major transactions. The filing supplies a useful boundary on the manifesto’s decentralization language: the proposed board process sits inside a company where its founder retains decisive voting control.
That does not show the board process will be ineffective. But the manifesto does not identify the directors involved, publish the safety criteria, or explain how outsiders could assess a release decision. It also leaves the terms of the free tier, extra-compute auction and private mode to future implementation.
The proposal is not an argument for unfettered release. Zuckerberg says wider access to cybersecurity capability can help users harden systems, while biological and chemical risks may need more coordination among governments and institutions. He proposes that frontier labs give government intermediate training checkpoints and technical staff to help secure critical systems before public release.
He also argues that U.S. policy should not slow American model releases, warning that even a month can matter in competition with foreign labs. The same statement calls for faster construction of energy, data-center and silicon infrastructure, and for continued export controls. The result is a specific allocation of roles: users would receive personal agents; government would receive early technical access; and labs would retain the responsibility to train, deploy and, under Meta’s proposal, decide what is released.
Company-reported capital expenditures were $72.22 billion in 2025; Meta guided to a $115–135 billion range for 2026. Source: Meta Investor Relations.
The financial scale helps explain why the details of access cannot be treated as an afterthought. Meta reported $72.22 billion in 2025 capital expenditures, including principal payments on finance leases. For 2026, it guided to $115 billion to $135 billion on the same basis, with the increase driven by investment in Meta Superintelligence Labs and the core business. That range is about 59% to 87% above the 2025 figure, based on the company’s stated figures in its earnings release.
Those figures are guidance, not a completed buildout. Still, they show that an individual-facing agent service would rest on a capital-intensive network owned or contracted by a small group of infrastructure operators. A paid path to more compute is therefore central to the offer: it is part of how scarce capacity would be allocated.
Meta pairs that construction program with a $1 billion fund for U.S. communities where it owns and operates data centers, supporting teachers, first responders, and energy and water infrastructure. The company also says it aims to avoid raising local energy prices and to restore more water than it uses. These are Meta commitments, not independent measures of completed local outcomes.
Zuckerberg’s track record makes the distinction between an objective and a demonstrated service important. In 2021, he recast Facebook as Meta and described the metaverse as the successor to the mobile internet, projecting a large-scale creator and commerce ecosystem. A retrospective assessment says the metaverse has remained well below that projected scale and that Reality Labs has accumulated losses in the tens of billions of dollars. That history does not determine the outcome of personal agents; it does argue against treating the manifesto’s forecasts as results.
The remaining question is not whether wider AI access could be valuable. It is whether Meta can show, in products and governance documents, that its model of access gives users meaningful choices rather than a broader route into a centrally operated service.
The announcement points to several testable next steps:
Until those details arrive, the manifesto is best understood as Meta’s case for a particular AI future: one where access is broad, but the infrastructure, defaults and initial rules of that access remain in the hands of the company building it.
Keel Infrastructure has completed its U.S. bitcoin-mining shutdown and is developing former power sites for high-performance computing. But its filing says none of the three priority U.S. sites had begun HPC operations or recorded related revenue as of Aug. 7, leaving contracts, funding and execution—not the shutdown itself—as the next test.
Bitdeer’s conditional lease for its Tydal, Norway data-centre conversion is material because the company plans to move 225 megawatts of the former mining site to colocation. But public accounts describe different contract values, terms and counterparties, leaving the customer, funding and revenue implications unresolved.
Nvidia is reportedly discussing an AI-infrastructure funding package of up to $500 billion with large private-capital firms. The talks underline the need to finance chips, data centres and power, but no commitments or risk-sharing terms have been disclosed.
Micron has signed 16 multiyear strategic customer agreements that commit buyers to specified memory volumes and, in many cases, pricing ranges. The contracts support its expansion plans and reduce risk for a portion of the business, but they do not yet make most of Micron’s revenue contractual or settle whether demand will outlast the supply shortage.
Anthropic, Macquarie Asset Management and GIC have formed Theseus Infrastructure to develop, operate and lease dedicated data centers to Anthropic. The investors will own the platform and provide most project equity, but the companies have not disclosed sites, capacity, costs or a delivery timetable—and Anthropic already has a separate $50 billion infrastructure plan with Fluidstack.
Sen. Bernie Sanders has asked OpenAI, Anthropic and Meta to pause AI development, invoking their conditional safety commitments. The request is not a pause order, but it extends his earlier push to halt new AI data-center construction pending safeguards.
Intel has proposed a $15 billion underwritten common-stock offering, with a further $2.25 billion underwriter option. The gross base deal equals about half of Intel’s June cash and short-term investments, yet the preliminary prospectus leaves the price, share count, net proceeds and allocation of capital unspecified.
Archer has agreed to acquire Boeing subsidiaries Wisk Aero, Insitu and SkyGrid in a stock-and-warrant transaction that keeps Boeing economically and technologically involved. Insitu gives Archer an operating defense business; the deal leaves unresolved how Archer will fund and organize the autonomous Wisk program alongside its piloted Midnight aircraft.
The police watchdog is investigating an unnamed senior Derbyshire detective over alleged AI use in decision logs—records that may be disclosed in court—as ministers press ahead with a £75 million national police-AI programme.
Industries that buy memory chips say tighter supply and higher prices are threatening products from network equipment to medical devices. Their competing requests to Washington—add capacity, reserve supply for domestic industry, or keep Chinese suppliers out—leave Commerce with a policy conflict, not a settled response.
Moore Threads has approved a plan to issue H shares and seek a Hong Kong main-board listing after first-half revenue rose 147%. The proposal has no disclosed size, price, timetable or use of proceeds, leaving investors to weigh a fast-growing cloud-products business against continuing losses, cash use and GPU competition.
A media tally says at least 37 people have been arrested in U.S. data-center protests in 2026, though it is not an official database. Two local cases show a narrower, documented conflict: who can speak at hearings over proposed infrastructure, under what rules, and with what recourse when those rules are enforced.
Meta has released Apache-licensed weights for Muse Glimmer, a 30-billion-parameter agentic model intended for local deployment. The hardware and benchmark claims make the proposition concrete, but they are Meta's measurements, and the harder test is whether developers can safely operate the tools around it.
TSMC reported NT$467.58 billion in July revenue, up 44.7% from a year earlier and described by contemporaneous reports as a monthly record. The total supports its bullish AI-and-HPC outlook, but the disclosure does not break out the customers, products or AI contribution behind the increase.
South Australia plans to begin a royal commission on artificial intelligence on 1 October and report by 1 July 2027. Its immediate test is whether the government can define a focused remit on AI use, work, public services and infrastructure consequences before federal standards move ahead.
Meta has seeded a $1 billion fund for U.S. communities where it owns and operates data centers. The commitment places community benefits at the center of Mark Zuckerberg’s AI expansion case, but the company has not disclosed the timing, allocation rules or reporting that would let a host community judge its bargain.
California’s SB 903 would stop companion chatbots from being presented as psychotherapy and set consent, clinician-review and accountability rules for specified AI uses in mental-health care. The bill is not a blanket ban on AI support; its unresolved challenge is where clinical assistance becomes clinical judgment.
Microsoft’s latest in-house text-to-image model is second in Arena’s August 10 ranking, behind OpenAI’s GPT-Image-2. The result is an early competitive signal from a live head-to-head-vote leaderboard, while Microsoft’s promised reference, grounding and control features—and their wider rollout—remain to be detailed.
ChatGPT search can surface available times from OpenTable, Resy and Yelp and complete some reservations. The rollout makes ChatGPT a new place to start a booking, but coverage, confirmation and every post-booking change remain with the reservation providers.
Yang Qi, Game Science’s co-founder and Black Myth art director, says the team will keep AI-generated content out of Black Myth: Zhong Kui’s design and asset-production stages. The Weibo response is a limited production choice for an early project, not a disclosed company-wide AI ban or a release-date update.