ETH Zurich produces provably perfect random numbers for the first time using a Bell test
A team led by Renato Renner and Andreas Wallraff reports in Nature a randomness-amplification procedure using two superconducting qubits linked by a thirty-metre cold tunnel, the first laboratory certificate for perfect randomness.
- ETH Zurich has reported the first certified procedure for generating perfectly random numbers.
- The experiment uses two superconducting qubits joined by a thirty-metre cold tunnel.
- A high-fidelity Bell test enables what the authors call randomness amplification.
- The result appeared in Nature on May 27, 2026.
A team of physicists at the Swiss Federal Institute of Technology in Zurich, led by Renato Renner and Andreas Wallraff, published in Nature on May 27 a paper demonstrating for the first time the production of provably perfect random numbers. The authors call the procedure randomness amplification, and it uses entangled superconducting qubits to remove systematic bias from existing generators.
The experimental setup consists of two superconducting chips cooled to temperatures near absolute zero. The chips are connected by a 30-metre tunnel, itself cooled, through which microwave photons travel to establish quantum entanglement. The configuration is for the first time stable enough that a Bell test delivers both high quality and a high rate of data generation.
What the problem with standard randomness is
Existing random-number generators, even those based on quantum effects such as the reflection of photons off a semi-transparent plate, are not fully immune to systematic errors. The result is a so-called bias, a tendency for some digits to appear slightly more often than others.
In cryptographic systems, even the smallest such deviation can be a crack. Banks, card processors and government communications security services have been developing protocols that assume absolute unpredictability of input numbers, something that has been difficult to confirm empirically until now.
Randomness-amplification methodology
The Wallraff team feeds imperfect random numbers from a standard generator into the system. These numbers are used to select measurement bases on the two entangled qubits. The Renner team then applies a specialised algorithm that transforms the measurement results into an output that can be certified as perfectly random.
The key is the quality of the Bell test, now run at a substantially higher sampling rate than earlier implementations:
- two superconducting qubits as transmon structures on separate chips
- microwave photons as the carriers of quantum entanglement
- a 30-metre tunnel that eliminates any classical communication as fast as light
- algorithmic post-processing that amplifies the randomness of the output
The resulting sequence of zeros and ones is now truly perfectly random, and we can even certify that. This is a field of research in which millimetre improvements in the quality of the Bell test carry a significance that is not obvious to the public.
Renato Renner stresses that certification is the key. Without it, the user has to trust the hardware manufacturer, and in security applications that is an unacceptable risk.
What the study does NOT show
The authors stress that the device, in its current form, is not commercially ready. The generation rate is sufficient for research purposes but is not competitive with existing industrial hardware generators that operate on millions of bits per second.
A second limitation concerns temperature and size. The experiment requires extreme cooling and a large laboratory infrastructure, so direct application in everyday security devices is not possible.
Andreas Wallraff warns that every modern random-number generator, even one based on quantum effects, is subject to systematic errors that the amplification approach can now eliminate, but only at the laboratory level. The move into engineered, mass-produced components is the next stage, which the authors estimate will take five to ten years.
Consequences for cyber security
As quantum attacks on classical cryptographic systems grow more capable, the quality of randomness sources becomes increasingly important. Post-quantum cryptographic algorithms currently being standardised at the United States National Institute of Standards and Technology require sources of randomness that we can be certain have no exploitable bias.
A certified source of randomness of the kind ETH demonstrates could over time become a standard component in server farms of government institutions and financial systems. The Mihajlo Pupin Institute in Belgrade has run a research programme in quantum communications for several years and could join the next generation of tests.
Will certified quantum randomness be built into everyday security systems by 2035?
Share your view: where do you see the biggest practical use of perfect random numbers, in cryptography, in simulations, or in testing AI systems? Tell us in the comments.
