A theoretical preprint posted to arXiv in April 2026 (arXiv:2604.11910) by Emanuele Polino and collaborators demonstrates that full network nonlocality can be achieved using only separable measurements augmented by bidirectional classical feedforward. This construction directly challenges the assumption threaded through nearly all prior network Bell experiments that a central party must perform an entangled measurement to certify that every independent source in the network is genuinely quantum.

This work is purely theoretical: the authors analytically derive a family of quantum correlations in a specific network topology (multiple independent entangled sources distributing states to three or more parties) and prove that these correlations violate a network Bell inequality even when all measurements are separable. There is no experiment, no dataset, and no sample size. As a preprint, it awaits peer review; its claims rest on mathematical proofs rather than laboratory data. Limitations explicitly acknowledged include idealized noise-free conditions and the requirement for precise classical communication channels, factors that will likely reduce the robustness of any future experiment.

Previous coverage in both specialist journals and popular science has consistently portrayed entangled measurements as indispensable. The seminal 2019 Nature paper by Renou et al. ('Quantum network nonlocality') used a star-shaped network with a central entangled projector to witness nonlocality across all sources simultaneously. Subsequent experimental realizations, including a 2022 Physical Review Letters demonstration of photonic network nonlocality, followed the same template and required joint measurements on multiple photons. What these accounts missed, and what Polino's team makes explicit, is that entanglement can be offloaded onto classical feedforward without sacrificing the ability to rule out local-realist explanations for every source.

Synthesizing this result with two related works sharpens the insight. First, a 2021 Quantum paper on 'minimal network nonclassicality' by Cavalcanti and colleagues introduced inequalities that prevent an observer from attributing the observed statistics to any fixed subset of classical sources. The new preprint shows the same separable-measurement strategy also violates these inequalities, unifying two seemingly distinct forms of network nonlocality under one practical roof. Second, building on the device-independent randomness certification framework developed by Acín, Pironio, and collaborators (Nature Communications, 2016), the authors quantify the extractable randomness. They prove that full-network-nonlocal correlations generated via separable measurements can certify roughly 1.2 bits of genuine randomness per run, only modestly lower than entangled-measurement protocols but far easier to implement.

The deeper pattern this reveals has been hiding in plain sight since the first network Bell papers: nonlocality in networks arises primarily from the independence of the sources rather than from the measurements. By allowing classical communication to coordinate separable local measurements, the protocol blurs the traditional quantum-classical boundary in a way that echoes the 'measurement-device-independent' quantum cryptography revolution of the early 2010s. It suggests that the 'spookiness' is located more in the correlations created by independent entangled sources than in the act of joint measurement itself.

This carries genuine foundational weight. Standard Bell nonlocality can still be explained by some nonlocal hidden-variable models if one allows measurement dependence. Network scenarios close that loophole by enforcing source independence. Demonstrating that even separable measurements suffice tightens the no-go theorem against classical explanations and forces us to confront whether retrocausality, superdeterminism, or a fundamental breakdown of locality best describes reality. At the same time, the practical payoff is immediate: entangled measurements (especially Bell-state measurements on photons) remain technologically demanding and loss-prone. Removing them lowers the bar for photonic, atomic, or superconducting quantum-network testbeds.

Future experimental groups will now race to implement this feedforward-separable protocol. If successful, it will not merely replicate existing network nonlocality demonstrations but do so with higher repetition rates and lower resource overhead, potentially accelerating the development of device-independent quantum key distribution across distributed nodes. The preprint therefore simultaneously advances quantum foundations, tightens the constraints on classical models of reality, and offers a more experimentally friendly route toward the quantum internet.