A Fraction-Of-A-Percent Flaw Is All It Takes to Fake Quantum Entanglement

Picture a physicist at a bench, running the test that nearly every quantum experiment leans on. Two particles, supposedly bound together; a screen that scores how tightly. The number arrives high and clean, a confident signature of deep entanglement. Everything checks out. Except the two particles share no connection whatsoever. The signal is a fake, and nothing in the readout will tell her so.

Quantum encryption rests on a promise you've probably absorbed from headlines: a message no one can secretly read, guaranteed not by clever math but by the laws of physics. The trick is to prepare two particles that share a single joint state, so fragile that anyone who tries to listen in has to touch them, and touching leaves a mark. But before you can trust that channel, you have to prove the particles really are entangled, and not an ordinary, unconnected pair dressed up to look the part. That proof is a lab test, and the test leans on instruments built, calibrated, and trusted by people.

Physicists have always known their instruments aren't perfect, and they've always assumed the imperfection is harmless: a small error makes a small mistake. Picture a ruler printed one percent too short. Every length comes out one percent off, the error staying in proportion to its cause. The idea is so obvious it's rarely said aloud.

The new paper shows that in exactly the systems built for secure communication, the proportion breaks. A familiar quantum bit has two possible states, like a coin. These high-dimensional systems have dozens: thirty-one distinct directions a measurement can point, or sixty-one. In that crowded space, two settings can sit a hair apart and still point at completely different things, like two compass needles that seem to agree while one reads north and the other east. And the test has no way to catch it, because the only thing it checks is average agreement, a single blurred score across all those directions. A measurement that matches the ideal across nearly every direction can be wrong where it counts, and the blur hides the difference. A tilt too small to see produces an answer that can be fabricated whole.

The damage scales with the crowding. With a deviation of 0.7 percent in a sixty-one-channel system, the model predicts a forged score certifying the equivalent of forty-seven independent entangled channels, all but indistinguishable from a genuinely well-entangled pair. To prove it wasn't just math, the researchers built the lie, and not with quantum particles. In the lab they ran a simpler version of the same attack, thirty-one channels instead of sixty-one, using ordinary beams of light steered by programmable holograms that dialed in a tiny, exact distortion. With the same 0.7 percent tweak, the standard test looked at a plain classical beam and certified eleven channels of entanglement that weren't there. Smaller still, a distortion of 0.23 percent, was enough to throw a false positive. The instrument was within spec. The test passed. Nothing was there.

The unsettling part isn't that quantum physics is broken; it isn't. The forgery takes a deliberate, finely tuned hand, an adversary engineering distortions too small to notice. It's a threat model, not an accident. The deeper trouble is what you can't do afterward: from the result alone, a true entanglement signal and a careful fake look exactly the same. There is no tell. The security proof at the heart of quantum cryptography assumes its instruments are sound, and that assumption was never guaranteed. Until the tests are rebuilt to survive small, malicious errors, the lock is only as trustworthy as the hands that made the key.