Two labs move quantum voting from theory to hardware, testing pools of up to four voters

The physical claim here is narrower than "secure voting", and worth stating exactly. Joey Marcellino's description of the problem is the useful one: the ideal anonymous ballot is everyone writing a vote and dropping it into a hat, which works only because everyone present can see the hat is a hat and not secretly recording which vote came from whom ^[17]. Move that online and you have an email server that promises to strip the sender's identity, where, as he puts it, he could simply be lying, and anyone who breaks into the server learns the same thing ^[11]. Protocols in this family answer that by encoding each vote so that an individual voter's output carries no information about that voter's choice ^[12]; the resource that makes the quantum version work is a GHZ state, a multipartite entangled state spanning three or more qubits ^[5]. What both papers report is that this ran without a trusted central election authority in the loop ^[2].

Scale is where the enthusiasm has to be metered. Between them, the two demonstrations counted twelve ballots ^[13]. The two designs pull in different directions: the Sorbonne candidate list is eight times Geneva's and its electorate twice as large ^[14], while Geneva's contribution was to simplify the 2022 protocol until an existing distributed polarization entanglement setup could be adapted to run it ^[9]. The Sorbonne group, for its part, reports developing the protocol to be more scalable and resource-efficient ^[7], and organised its eight voters as two pools of four ^[6].

What this does not tell you is how the resource grows. GHZ states span three or more qubits, and the reported pools are four voters wide, which implies the entangled state has to grow with the voting group ^[19]. Whether four was a protocol design choice or the largest state the photon source could hold at usable quality is not something the published account settles. Nor does the account carry a photon generation rate, a separation distance between voters, or an elapsed time per ballot ^[18], and those are the numbers that decide whether a protocol can work anywhere other than an optical table.

Nicolas Laurent-Puig ties the motivation to a concrete failure, citing "the recent Classical Voting event for the election in France, which was later shown to be insecure" ^[10]. The report does not identify that event further ^[16], which matters for reading the result, because the classical weakness being answered here is one specific weakness, trust in whoever aggregates the ballots, rather than the wider set of problems an election has to survive.