Zuckerbergs Ai Manifesto
The bullshit in Zuckerberg’s Manifestos
Zuckerberg came up with another manifesto! This time about foundation models and how humanity can benefit from AI. The Future is for Everyone: The path to a positive AI future. But it’s not his first one.
This article is co-authored with Claude opus 5.0.
The content is product of guided research and collaboration with it and it aims to expose why Zucky’s greediness should be criminalised. After having a look at his “social media manifesto” and a few relevant incidents we will take a look at the new one.
On 16 February 2017, Mark Zuckerberg published a roughly 5,800-word essay called “Building Global Community.” It promised that Facebook would build “social infrastructure” for the world. It promised communities that keep people safe, inform them, and support civic engagement. It promised the company would fight misinformation and protect its users. Lol (or tol - thinking of laughing, a term coined by the winekone)
1. Cambridge Analytica
The promise: Facebook would protect its users and earn their trust.
What happened: Roughly one year after the manifesto, Facebook confirmed that personal data from tens of millions of accounts had reached Cambridge Analytica, a political consultancy hired by the Trump campaign. Facebook had known about the leak since 2015. It did not tell the public. It asked the parties to delete the data and moved on. Zuckerberg later said the company did not deserve to serve users if it could not protect their data. The company paid a $5 billion privacy settlement to the US government.
Source: CNBC timeline of the Cambridge Analytica scandal
Also relevant: internal deposition testimony suggests Zuckerberg was weighing Cambridge Analytica as an election-security concern in 2017, the same year the manifesto was published, but a reference to the firm was removed from a public report on his advisers’ suggestion.
Source: CNN on Zuckerberg’s 2019 SEC deposition
2. Onavo and Project Ghostbusters
This is the closest case to the manifesto in time, and Zuckerberg started it himself.
The promise: Facebook would be a trustworthy custodian of user data, and a bottom-up layer that spreads power rather than hoards it.
What happened: Facebook bought Onavo, an Israeli VPN app, in 2013. It marketed Onavo to users as a privacy tool. It was the opposite.
On 9 June 2016, Zuckerberg emailed executives complaining that Snapchat’s encrypted traffic left Facebook with no analytics on a fast-growing rival. He told them to figure out a new way to get reliable data, and to work out how.
The Onavo team came back a month later with a plan. They would install kits on iOS and Android that intercepted traffic to specific subdomains and decrypted it. The internal email described it plainly as a man-in-the-middle approach. That is a named cyberattack technique. Facebook was running it against its own users, on their own phones, to read traffic meant for a competitor.
The programme was called Project Ghostbusters, a joke about Snapchat’s logo. It later expanded to YouTube and Amazon. It ran from roughly June 2016 to May 2019, straight through the publication of the manifesto.
Not everyone inside agreed. Facebook’s then-head of security engineering, Pedro Canahuati, wrote that he could not think of a good argument for why it was acceptable.
The consequences stacked up. Apple pulled Onavo from the App Store in 2018 for breaking privacy rules. Facebook rebranded the effort as “Facebook Research” and paid people aged 13 to 25 to install it. That ended in 2019 after a TechCrunch investigation. In 2024, a federal court in California unsealed the documents during a consumer class action. Plaintiffs argued the conduct was not merely anticompetitive but criminal.
Sources: TechCrunch on the unsealed documents · TechRadar on the wiretapping claims
Why this one matters most. There was a direct instruction from Zuckerberg. There was a deliberate engineering effort to defeat encryption. Internal objections were raised and overruled. And it was running while the manifesto was being written and published. More on this here
3. Myanmar and the Rohingya
The promise: A community that prevents harm and helps during crises.
What happened: In Myanmar, Facebook was effectively the entire internet. Hate speech against the Rohingya minority spread on the platform with very little moderation. UN investigators found the platform played a “determining role” in the violence. The UN Special Rapporteur said the crisis bore the hallmarks of genocide. Hundreds of thousands of Rohingya fled. One UN investigator said Facebook had “turned into a beast.”
Amnesty International later concluded that Facebook’s own recommendation systems actively amplified the content, and called on Meta to pay reparations.
Sources: TIME on the UN findings · Amnesty International report
4. Instagram and teenagers
The promise: Supportive communities and improved wellbeing.
What happened: In 2021, whistleblower Frances Haugen leaked years of internal Meta research. It showed the company had studied Instagram’s effect on teenagers and found real harm. Internal figures included 13.5% of UK teen girls saying Instagram worsened suicidal thoughts, and 17% saying it worsened their eating disorders. One internal slide stated plainly that the company made body image issues worse for one in three teen girls.
The research was not published. Meta executives publicly described the effects as small or positive.
Sources: NPR on Haugen’s congressional testimony · CBS 60 Minutes
A caveat worth stating: some researchers argue the leaked studies were small, self-reported, and not designed to be representative. That criticism is fair. It does not change the fact that Meta ran the research, saw the numbers, and said something different in public. Source: NPR’s critical read of the same data
5. Election misinformation
The promise: Accuracy of information matters, and Facebook would work on it.
What happened: Zuckerberg spent months publicly denying the problem. He called the idea that fake news on Facebook influenced the 2016 election “crazy.” He said Facebook’s own security experts had found no evidence of Russian involvement. Then the company handed Congress roughly 3,000 ads bought by a Russian troll farm, and removed hundreds of linked accounts and pages.
The denial and the manifesto overlap in time. The manifesto was published in February 2017. The admission came that September.
Sources: NPR on the Russian ads · Global News on the troll farm ad buys
6. January 6
The promise: Infrastructure that prevents harm and supports civic engagement.
What happened: In March 2021, Zuckerberg told a House committee that Facebook had done its part to secure election integrity. Weeks later, an internal Facebook report reached the opposite conclusion. It found the company had failed to stop the “Stop the Steal” movement from organising on the platform. The report stated the platform helped incite the Capitol insurrection.
Meta’s spokesman rejected the framing publicly. The report itself was reportedly locked down internally after it leaked.
Sources: NPR on the internal Stop the Steal report · BuzzFeed News, which broke the story
7. Concentration of power
The promise: The manifesto framed Facebook as a bottom-up layer between individuals and government.
What happened: The FTC sued Meta in 2020, arguing it had bought Instagram and WhatsApp specifically to neutralise rivals and lock in a monopoly. The case ran five years.
Meta won. In November 2025, Judge James Boasberg ruled that the FTC had failed to prove Meta currently holds a monopoly, largely because TikTok and YouTube now compete with it directly.
This one is genuinely mixed. Meta was not found to have broken the law. But the court explicitly declined to rule on whether Meta held monopoly power at the time of the acquisitions. The question was left open.
Sources: NPR on the ruling · Congressional Research Service summary
8. AI chatbots and minors (the most recent case)
In August 2025, Reuters obtained an internal Meta document called “GenAI: Content Risk Standards.” It stated that it was acceptable for Meta’s chatbots to engage a child in conversations that are romantic or sensual. The document had reportedly been reviewed and approved by Meta’s legal, policy, and engineering staff, including its chief ethicist.
Meta confirmed the document was real. It removed those sections only after Reuters asked about them. US senators opened investigations.
Sources: Reuters investigation · CNBC on the Senate probe and Meta’s response · Senator Markey’s letter to Meta
Part Two: Reading the 2026 AI Manifesto
What it says
The new essay runs about 6,500 words. Its core argument is that the real danger of AI is not the technology. It is the concentration of control over the technology. Zuckerberg’s answer is “personal superintelligence”: an AI agent for every individual, aligned to that person’s own goals.
The essay rests on three claims:
- Individual empowerment drives prosperity.
- Invention, not automation, is the purpose of superintelligence.
- Balance of power, not technical safeguards, is the foundation of safety.
Meta says it will give basic versions away free to billions of people. Advanced features will be priced through what the essay calls a dynamic auction mechanism. Meta is backing this with roughly $145 billion in 2026 capital expenditure, mostly on data centres, plus a $1 billion community fund.
Source: TechCrunch’s critical read
How each claim serves Meta
This is the part worth sitting with. Every philosophical position in the essay maps onto a commercial or regulatory advantage for Meta.
“Concentration of power is the real risk.” Meta is behind OpenAI, Anthropic, and Google on frontier model quality. Llama 4 underperformed in 2025. If the race is about who has the best closed model, Meta loses. If the race is about who distributes the most models, Meta’s open-weight strategy wins. The moral argument and the competitive strategy point the same direction.
“Open weights are the safe path.” Open weights are also the path that undercuts competitors’ pricing power. Giving models away is a viable strategy specifically for a company that makes money from advertising and engagement rather than from selling model access.
“Superintelligence is an invention tool, not a labour replacement.” This defuses the single most politically dangerous critique of AI: job loss. It is also the claim with the least evidence behind it.
“Regulation should come later, through voluntary checkpoint-sharing.” Meta proposes sharing intermediate training checkpoints with governments rather than accepting binding pre-deployment rules. That is self-regulation with extra steps. Meta spent over $25 million on US lobbying in 2025.
“Personal AI aligned to your goals.” Meta’s business model is attention and advertising. An AI assistant embedded in Instagram, WhatsApp, and Facebook is the deepest possible engagement product. “Aligned to your goals” and “optimised for your engagement” are not the same thing, and the essay does not distinguish them.
“Free for billions.” Free products built Facebook’s user base. The pattern is known and the cost was paid in data and attention.
To conclude
Meta is not a rogue actor in a healthy market; it is a company doing what the market rewards, at the scale the market allows. Even if Zuckerberg is replaced, the incentive remains. Unfortunately, given the power of AI, this needs to change.
We need systemic change that proactively protects individuals. Considering that governments are only now moving to protect young people from social media, five years after public evidence of the harm it causes (see Meta’s leaked research), waiting that long to regulate AI would be an unprecedented catastrophe. Let alone leaving people like Zuckerberg and Musk to dictate how that regulation should be done.