Three Manifestos on Who Should Govern AI in Less Than 30 Days. Coincidence?
They’re Not Debating the Future of AI. They’re Negotiating Who Gets Permission to Govern It.
Zuckerberg, Gates, and Amodei published three manifestos on AI within just 33 days. The most disturbing theory isn’t that they coordinated their message. It’s that the people who control the infrastructure have realized that the time has come to fight over who gets to write the rules of the system.
There is a more uncomfortable hypothesis here, but we need to distinguish between a plausible investigative theory and the idea that “they all sat in a room and planned this,” for which I found no evidence.
The truly relevant coincidence isn’t simply that the three manifestos appeared within 33 days. It’s that all three shift the conversation away from what AI can do and toward who has authority over AI: who gets access, who gets to impose limits, who gets to audit it, and who gets to decide when continuing becomes too dangerous.
Zuckerberg literally opens his manifesto by asking who will have access to superintelligence and what it will be directed toward. Amodei proposes independent evaluation, coordination among labs, and mechanisms for slowing development.
The More Interesting “Conspiracy Theory”
It wouldn’t be:
“Zuckerberg, Gates, and Amodei coordinated their manifestos.”
It would be something structurally more plausible:
The industry has realized that we are approaching the moment when governments may begin claiming authority over AI. These manifestos are an attempt to establish, in advance, which models of governance will be considered acceptable — while keeping frontier AI labs inside the circle that gets to decide.
In other words, we may be watching Big Tech evolve into a kind of quasi-sovereign infrastructure.
Look at the hidden detail.
Zuckerberg isn’t simply saying, “let the market decide.” He proposes deep cooperation between AI labs and government, including giving governments access to intermediate checkpoints of advanced models and technical personnel. He also proposes independent governance over deployment decisions.
Amodei approaches the issue from the opposite direction: “this is becoming too dangerous.” But his solution also involves labs + experts + auditors + governments coordinating the frontier.
They disagree about speed and distribution.
But they are beginning to converge on something much bigger:
the AI frontier will not be governed like an ordinary app.
And this is already moving beyond theory. OpenAI, Anthropic, and Google DeepMind have been discussing coordination around safety.
That is where the timing becomes politically interesting.
Maybe the Real Product of These Manifestos Is Legitimacy
During the first phase of the race, companies needed:
capital → chips → researchers → data → energy.
In the next phase, they will need another resource:
social license to possess systems more powerful than public institutions can build themselves.
And social license cannot be bought from Nvidia.
It has to be manufactured through narrative.
So three extraordinarily useful narratives emerge:
Zuckerberg: “We are not accumulating power. We are democratizing power.”
Gates: “We need to preserve the human.”
Amodei: “We cannot let this escape our control.”
Look at the result.
The companies stop appearing merely as the creators of the problem and begin occupying the position of architects of the solution.
That symbolic shift is worth billions.
Now Comes the Truly Uncomfortable Hypothesis
Maybe the real battle isn’t about who will have the best AI.
Maybe a different battle is beginning:
who will be authorized to possess superintelligence.
Imagine that, after serious incidents, Congress concludes:
“Models above a certain capability threshold must be audited, certified, monitored, protected against cyberattacks, equipped with shutdown mechanisms, and required to comply with international protocols.”
Sounds reasonable.
Now ask:
how many companies can actually comply?
Meta.
Google.
Microsoft/OpenAI.
Anthropic.
xAI.
Maybe a few others.
Regulation designed to prevent catastrophe could inadvertently create the largest competitive moat in the history of technology.
And the American antitrust regulator has already noticed this risk. The FTC chair has warned that requests for antitrust exemptions and new regulations could end up creating barriers that protect established companies.
That changes the entire interpretation.
Because two things can be true at the same time:
Advanced AI may genuinely require extraordinary controls.
and
Those controls may extraordinarily consolidate the position of those who already dominate AI.
We do not have to choose between “existential risk” and “regulatory capture.”
One can produce the other.
The More Dissident Scenario: The Birth of the “Frontier Club”
My investigative hypothesis would be this.
Not a secret cartel.
Something more sophisticated — and completely public:
an oligopoly legitimized by safety.
The process could unfold like this:
AI capabilities increase → incidents increase → public fear grows → CEOs demand governance → governments need technical expertise → they turn to the labs themselves → labs help write the standards → standards become extremely expensive → smaller companies cannot comply → only a handful of labs remain authorized to operate at the frontier.
The result:
power becomes concentrated precisely in the name of preventing the dangerous concentration of power.
That would be the perfect paradox.
And it requires no conspiracy.
Every actor can be behaving rationally and even sincerely.
There Is an Even Bigger Clue in Zuckerberg’s Manifesto
He writes something extraordinarily explicit: Meta considers it important that multiple frontier labs exist so that their systems and values can counterbalance one another.
Read that concept again.
Not billions of companies.
Not necessarily governments building their own systems.
“Multiple frontier labs.”
Amodei wants coordination among those labs.
OpenAI, Anthropic, and Google are already discussing coordination on safety.
Microsoft has just published its own kind of “constitution” for keeping AI under human control.
So the question begins to shift from:
“Who will win the AI race?”
to:
“Who will be admitted into the small group authorized to keep racing?”
That is much bigger.
And Gates Occupies a Curious Position in This Architecture
He does not need to win the model race to participate in defining the moral boundaries of the era.
His Human Reserved proposal introduces another kind of authority:
not merely who controls the machine, but who determines where the machine will not be allowed to enter.
Education.
Healthcare.
Caregiving.
Work.
Human decision-making.
This means we may be watching three forms of power emerge simultaneously:
computational power: who possesses intelligence;
regulatory power: who decides how much intelligence is permitted;
normative power: who decides where artificial intelligence should replace humans.
And the manifestos are competing precisely over these three positions.
What Almost No One Is Seeing
Perhaps we are making a historic mistake by asking:
“Who will govern AI?”
Because that question assumes AI is the thing being governed.
If agents come to control intellectual work, programming, scientific research, communication, digital infrastructure, cyber defense, and a growing share of economic administration, then governing AI means governing a layer through which other things will be governed.
It is the difference between controlling a factory and controlling electricity.
Between owning a newspaper and owning the infrastructure of the internet.
In that scenario, the major AI labs stop looking like ordinary companies.
They begin acquiring the characteristics of private constitutional institutions.
They establish:
what the system can do;
who can access it;
which behaviors are prohibited;
which values enter its training;
when a capability becomes dangerous;
when a model can be released;
who gets to audit it;
and, potentially, which states are allowed to receive it.
That is the layer I would investigate.
Not “the three of them coordinated their manifestos.”
That is too small.
The much more disturbing hypothesis is:
We are witnessing the informal negotiation of a new division of sovereignty between states and AI companies, and these manifestos are among the first public documents in which the people who control the infrastructure are attempting to define the constitutional principles of that division.
If that is correct, the 33 days stop looking like an editorial coincidence.
They begin to mark a moment when the owners of the technology realized that whoever builds the intelligence will not necessarily govern the intelligence — unless they also manage to write the rules by which it will be governed.
And this is the question I would put at the center of the article:
What if these manifestos are not trying to convince us about the future of AI, but trying to secure a place for their authors in the government of that future?
The Three Manifestos
Mark Zuckerberg — The Future Is for Everyone
Meta, August 10, 2026
https://www.meta.com/thefutureisforeveryone/
Bill Gates — The Turbulent AI Era Is Here
Gates Notes, August 26, 2026
https://www.gatesnotes.com/
Dario Amodei — We Must Pace the Frontier
September 12, 2026
https://darioamodei.com/post/we-must-pace-the-frontier
Follow Tech Gossip™ to see what happens when technology stops being merely a product and starts competing for sovereignty, language, and power.
#TechGossip #ArtificialIntelligence #AIGovernance #Superintelligence #BigTech #Meta #Anthropic #OpenAI #FutureOfAI #AISafety #TechPower #DigitalSovereignty #AlgorithmicPower #AIGovernance #Dissident