Quantum computers can solve problems far too complex for normal computers, at least in theory. That’s why research teams around the globe have strived to build them for decades. But this extraordinary power raises a troubling question: How will we know whether a quantum computer’s results are true if there is no way to check them? The answer, scientists now reveal, is that a simple quantum computer—whose results humans can verify—can in turn check the results of other dramatically more powerful quantum machines.

Quantum computers rely on odd behavior of quantum mechanics in which atoms and other particles can seemingly exist in two or more places at once, or become “entangled” with partners, meaning they can instantaneously influence each other regardless of distance. Whereas classical computers symbolize data as bits—a series of ones and zeroes that they express by flicking switchlike transistors either on or off—quantum computers use quantum bits (qubits) that can essentially be on and off at the same time, or in any on/off combination, such as 32% on and 68% off.

Because each qubit can embody so many different states, quantum computers could compute certain classes of problems dramatically faster than regular computers by running through every combination of possibilities at once. For instance, a quantum computer with 300 qubits could perform more calculations in an instant than there are atoms in the universe.

Currently, all quantum computers involve only a few qubits “and thus can be easily verified by a classical computer, or on a piece of paper,” says quantum physicist Philip Walther of the University of Vienna. But their capabilities could outstrip conventional computers “in the not-so-far future,” he warns, which raises the verification problem.

Scientists have suggested a few ways out of this conundrum that would involve computers with large numbers of qubits or two entangled quantum computers. But these still lie outside the reach of present technology.

Now, quantum physicist Stefanie Barz at the University of Vienna, along with Walther and their colleagues, has a new strategy for verification. It relies on a technique known as blind quantum computing, an idea which they first demonstrated in a 2012 Science paper. A quantum computer receives qubits and completes a task with them, but it remains blind to what the input and output were, and even what computation it performed.

To test a machine’s accuracy, the researchers peppered a computing task with “traps”—short intermediate calculations to which the user knows the result in advance. “In case the quantum computer does not do its job properly, the trap delivers a result that differs from the expected one,” Walther explains. These traps allow the user to recognize when the quantum computer is inaccurate, the researchers report online today in Nature Physics. The results show experimentally that one quantum computer can verify the results of another, and that theoretically any size of quantum computer can verify any other, Walther says.

The existence of undetectable errors will depend on the particular quantum computer and the computation it carries out. Still, the more traps users build into the tasks, the better they can ensure the quantum computer they test is computing accurately. “The test is designed in such a way that the quantum computer cannot distinguish the trap from its normal tasks,” Walther says.

The researchers used a 4-qubit quantum computer as the verifier, but any size will do, and the more qubits the better, Walther notes. The technique is scalable, so it could be used even on computers with hundreds of qubits, he says, and it can be applied to any of the many existing quantum computing platforms.

“Like almost all current quantum computing experiments, this currently has the status of a fun demonstration proof of concept, rather than anything that’s directly useful yet,” says theoretical computer scientist Scott Aaronson at the Massachusetts Institute of Technology in Cambridge. But that doesn’t detract from the importance of these demonstrations, he adds. “I’m very happy that they’re done, as they’re necessary first steps if we’re ever going to have useful quantum computers.”