Future 4 November 2026 4 MIN READ
Belgium 2003: The 4,096 Extra Votes
One candidate, 4,096 phantom votes and a bit flip. What the official report concluded, and what later retellings added about cosmic rays.
THE SHORT ANSWER
In 2003 a Belgian polling station counted exactly 4,096 votes too many for one candidate, a power of two. Official experts concluded that a spontaneous bit flip in memory was the most probable cause, by elimination. Their report never mentions cosmic rays, which came from later interpretations.
On this page
An impossible total
On the night of the Belgian federal election of 18 May 2003, officials in the Brussels district of Schaerbeek noticed something that arithmetic does not allow: one candidate had more preference votes than the whole list they stood on. The college of experts that oversaw the electronic voting system was told at 23:30.
The investigation started the next day, with the canton president, the interior ministry, the voting-system vendor Steria and the experts. They traced the problem to one polling station and recounted its ballot box from the magnetic cards. The new count showed no error. The gap between the two counts had been exactly 4,096 votes.
Why 4,096 stands out
The number is 2 to the power of 12. The experts noted in their report that a gap of exactly 4,096 can come from a single binary digit in a counter flipping from 0 to 1: the 13th position, counting from the right. A software bug would normally produce an arbitrary number. A one-bit hardware slip produces a power of two.
What the official report says
The experts met the vendor and an interior ministry representative on 22 May and concluded that the error could “probably” be put down to a spontaneous inversion of a bit in the computer’s memory. In the final report the wording is “très probablement”: most probably a spontaneous, random inversion of a binary position, a phenomenon the report calls “abundantly described in the specialised literature”.
The reasoning was by elimination. Tests on the polling-station president’s computer on 22 and 23 May, including memory tests and a simulated vote for the same candidate, found no hardware fault and could not reproduce the incident. The source code was analysed and no instruction that could have caused it was found. The experts judged the incident most likely isolated and without effect on the election result.
The report also records a weakness in the design: the software did not count twice or cross-check, so a flipped counter would go unnoticed unless it produced an impossible total, as it did here.
Where the cosmic rays come from
The popular version of the story says a cosmic ray hit the memory chip. Here the documents are more careful than the retelling. The official report, as we read it, never mentions cosmic rays and gives no evidence about what particle, if any, was involved. Contemporary coverage in 2003 and 2004 summarised it the same way.
The cosmic-ray link is a later interpretation. In February 2017, the Vanderbilt University engineer Bharat Bhuva presented the Schaerbeek case at a meeting of the American Association for the Advancement of Science as a bit flip that added 4,096 votes, a firmer wording than the Belgian experts’ own. Critics point out that the step from “unexplained bit flip” to “cosmic ray” is not supported by the evidence, and David Glaude of the Belgian e-voting watchdog PourEVA told Radiolab that the experts blamed a bit flip because they had found no bug.
The physics is not in doubt. IBM researchers showed in 1981 that cosmic rays at sea level can cause soft errors in memory, and they estimated that an 8-megabyte memory made of the diodes they studied would suffer about one soft error per day. That shows the mechanism is real. It does not show that it happened in Schaerbeek.
A note on the sources
The report is an official document of the statutory expert college, but because it was not available on official sites, the copy we read is a digitised transcription published by PourEVA, an organisation sceptical of electronic voting. The wording we quote is the French original.
Why it matters
Whatever flipped the bit, the case shows how an unprotected computer can quietly turn one wrong digit into thousands of votes, and that the error was caught only because the result was arithmetically impossible. A smaller error, a flip that left the totals plausible, would not have been noticed.
Our take: documented, probable, not proven. A single-bit error is the experts’ most probable explanation. The cosmic ray is an attractive story that the documents do not support.
Do you agree?
Take it to Chevlume: write your view, add your sources and let others answer. The closed alpha is by invitation.
Keep reading
Knight Capital: 45 Minutes, $460 Million
A forgotten code path on one of eight servers cost a trading firm $460 million in 45 minutes. What the SEC found about deployment and controls.
FUTURE / EXPLAINERWhat Quantum Computers Can and Cannot Do
Qubits, errors, advantage claims and the threat to encryption, with every result labelled as demonstrated, claimed by a company or projected.
BIOTECH / DISCOVERYTherac-25: When Software Was the Only Safety Net
Six overdoses between 1985 and 1987 from a radiation machine whose safety rested on software. What the Leveson and Turner investigation found.
Sources
- 18/05/2003: Élections fédérales 2003 : Rapport intégral du Collège des experts (§5.3.7 L’incident de Schaerbeek) · Collège d’experts (report to the Belgian Minister of the Interior and the federal Parliament), digitised by PourEVA · Accessed
- Irish Labour Party wants to stop e-voting (EDRi-gram no. 23) · European Digital Rights (EDRi) · Accessed
- The RISKS Digest, Volume 23 Issue 47 · ACM Committee on Computers and Public Policy / Newcastle Univ. (catless.ncl.ac.uk) · Accessed
- Alien particles from outer space are wreaking low-grade havoc on personal electronic devices · Vanderbilt University News · Accessed
- Cosmic Rays Did Not Change Election Results · Paco Hope (personal blog; author is a software-testing / security professional) · Accessed
- Bit Flip - Transcript · Radiolab (WNYC Studios) · Accessed
- The effect of sea level cosmic rays on electronic devices (J.F. Ziegler, W.A. Lanford) · IBM Research / Journal of Applied Physics · Accessed
Published
Discuss on X