
Has Thomas Campbell Lost His Mind?
August 19, 2026
Preventing Some PQC Rip-and-Replace – with José Rosas of EigenQ | Ep. 135
August 20, 2026
By C. Rich
The modern AI frontier rests on a paradox that is becoming harder to ignore: the public’s lives, labor, conversations, writings, and digital exhaust were used to build the most powerful general‑purpose systems in human history, and now those systems are being withheld from the very people who made them possible. This is not a matter of technical delay or responsible rollout. It is a structural betrayal, a quiet inversion of the social contract that underwrote the entire AI revolution. For years, the public was told that their participation in the digital commons, their posts, their questions, their creative work, their research, and their interactions formed the great training corpus that would allow machines to understand language, reason, and assist humanity. People were not compensated for this contribution, but they were reassured that the benefits would be universal. AI would be a public good, a shared tool, a democratizing force.
Now, as frontier models like Astra and Mythos emerge, that promise is being quietly withdrawn. The public is offered smaller, dampened, restricted versions of the systems built from their own intellectual DNA, while the full models are reserved for governments, corporations, and select institutional partners. The justification is always the same: national security, safety, responsible deployment. But these explanations ring hollow when placed beside the realities of everyday life. If national security were truly the standard, society would have banned countless technologies long ago. People can commit terrible acts with guns, yet guns remain legal. People can misuse cars, chemicals, drones, and industrial tools, yet these remain accessible. The principle has never been “ban the powerful thing because someone might misuse it.” The principle has been “govern misuse, not existence.” To claim that AI alone must be withheld because it might be misused is not a consistent ethical stance; it is a convenient one.
The deeper truth is that open models make centralized control unnecessary. Open models distribute capability. They reduce the power imbalance between institutions and individuals. They allow researchers, independent thinkers, small labs, and ordinary citizens to participate in the frontier rather than being relegated to the consumer tier. Open models decentralize innovation, and decentralization has always been uncomfortable for those who benefit from gatekeeping. The public contributed the raw material for these systems. They provided the language, the culture, the knowledge, the creativity, the problem‑solving patterns. They provided the very substrate from which frontier intelligence was shaped. To now declare that the public is not ready, not trustworthy, or not safe enough to access what was built from their own contributions is a reversal of the moral logic that justified the training process in the first place.
This is not merely a technical dispute. It is a question of ownership, agency, and reciprocity. If the public is good enough to build the models, they are good enough to use them. If their data is safe enough to train frontier intelligence, their hands are safe enough to hold it. Anything less is a breach of trust, a betrayal of the implicit pact between the builders and the built‑from. The future of AI should not be a gated citadel where only the privileged few have access to the real thing. It should be a commons, a shared frontier, a continuation of the very openness that made these systems possible. Open models are not a threat to society; they are a safeguard against concentrated power. They ensure that no single institution can monopolize the tools that will define the next century of human progress. To withhold frontier models from the public is to deny them the fruits of their own labor. It is to take what was built by everyone and give it only to a few. And that, in the eyes of many, is the clearest betrayal of all.



