A Quantum Computer Made Random Numbers. The Wild Part Was Checking Them.

Suppose a stranger online offers to flip a coin for you. You receive heads or tails, but how do you know they actually flipped anything? A quantum-computing experiment published in March 2025 tackled a much more sophisticated version of that trust problem: producing remotely generated randomness that could be checked.

The team used Quantinuum’s 56-qubit H2-1 computer as an untrusted server. Researchers sent it carefully chosen quantum circuits, essentially small programs, and required quick responses. The answers carried statistical features associated with those circuits, making the task much harder to fake with a conventional computer under the experiment’s assumptions.

Then came the expensive checking. As JPMorganChase’s research team explains, four supercomputers helped verify the results, including Oak Ridge’s Frontier. The researchers certified 71,313 bits of entropy: a quantified amount of unpredictability in the output. The achievement concerned evidence about the randomness, not merely producing a string that looked jumbled.

The qualification matters. The paper’s guarantee covered a restricted class of potential adversaries and depended on additional assumptions. This was not an unconditional proof against every possible attacker. Ordinary computers also already obtain randomness from physical sources; the experiment explored how to reduce trust in a remote provider.

That makes randomness an intriguing quantum-computing application, with potential uses wherever people need unpredictable choices they can scrutinize. It also shows how demanding a useful demonstration can be: a quantum machine supplied answers, while substantial classical computing did the checking. A digital coin toss became an unusually serious collaboration.