$QUBIC does not need slogans right now. The news speaks for itself. Last night, a single government decision was enough to cut off worldwide access to two of the most powerful artificial intelligences ever built.
Let us go back to the facts, calmly.
On June 12, the US Department of Commerce ordered Anthropic to suspend access to its two most advanced models, Claude Fable 5 and Claude Mythos 5, in the name of export control and national security. The directive aimed to prevent any foreign national from accessing them. Since partial enforcement was impossible, Anthropic had to disable these models for absolutely everyone, everywhere in the world.
Let us be precise, because precision is a matter of respect. This is not a ban for being dangerous. It is not permanent. Anthropic is contesting the decision and hopes to restore access. And it is a serious company, sincere in its commitment to AI safety. There is no reason here to rejoice, neither over the fate of a team, nor over that of its users deprived of a tool.
But there is a lesson, and it is immense.
In one evening, through a single administrative letter, access to a technology used by millions of people was cut off. Not because of a technical flaw. Not because of an outage. Because of a political decision, taken within one jurisdiction, applicable to a centralized actor. When an intelligence lives on the private servers of a company subject to a state, it can vanish overnight, for reasons that have nothing to do with its quality or usefulness.
This is exactly what the cypherpunks anticipated thirty years ago. Concentrated power always ends up depending on the will of a few. Yesterday it was money and privacy. Today it is intelligence.
This is why decentralized AI is not an ideological whim. It is a structural necessity. An intelligence spread across hundreds of thousands of machines, with no single point of shutdown, cannot be turned off by a letter. Not because it defies the law, but because it offers no single lever to press.
An essential point of honesty. Qubic and Aigarth are not there yet. Decentralizing intelligence is a work in progress, not an achievement. No one should claim a fully functional alternative already exists today. But the direction has never been as clearly justified as it was last night.
The question is no longer whether decentralized AI makes sense. The question is how many shutdowns it will take before we understand why it matters.
Tick after tick. Computor after Computor. You do not build an alternative for the day when all goes well. You build it for the day it becomes indispensable.