Dominion v. Fox: A Claim About Software Against Internal Messages
Wilmington, Delaware, 18 April 2023. A jury had been selected that morning, and opening statements were next. Instead, the judge announced the settlement from the bench. The number: $787.5 million.
How does a case get there before a single opening statement? Two sets of material set against each other, and one ruling that arrived before the trial did.
“The machine switched the votes” is not an opinion. It is a statement about how software behaves, so it can be tested: run it, read the code, compare it to the logs. One set was a claim about software. The other was material generated by digital tools, which Dominion argued showed what the speakers actually believed.
What actually happened
Dominion Voting Systems sued Fox News Network in the Delaware Superior Court, a state trial court. The suit concerned broadcasts that claimed Dominion’s voting machines helped tilt the 2020 U.S. presidential election. Dominion sought $1.6 billion in damages.
Then came 31 March 2023. Judge Eric M. Davis issued an opinion of about 80 pages on the summary-judgment motions. He denied Fox’s motions and granted Dominion’s in part. Among other things, he wrote that the evidence developed in the proceeding made it “CRYSTAL clear that none of the Statements relating to Dominion about the 2020 election are true”. In practice, the jury would not weigh whether the statements were accurate. The judge would instruct them that they were false.
Davis also rejected Fox’s “neutral reporting” defence, that it was shielded for neutrally conveying what newsworthy people said. Per NPR, he also found the statements defamatory per se.
What was left for trial? Per the opinion, three jury questions: whether Fox Corporation published the statements, whether there was actual malice (knowing falsity or reckless disregard for the truth), and whether Dominion suffered damages.
So weeks before trial, a court had decided, on the case record, that the claims about Dominion’s machines were false. What remained was a question about the speakers: what did they know and believe?
The settlement that followed came to $787.5 million, about half of what Dominion sought. Fox said afterward: “We acknowledge the Court’s rulings finding certain claims about Dominion to be false.” Per Reuters, legal experts called it the largest settlement struck by an American media company. There was no verdict; the case settled before opening statements. From here on I rely only on the allegations, the court’s rulings, the material before it, and reports of the settlement.
The artifact: a claim about software, and messages generated by software
Two artifacts, and the point is the tension between them.
The first is the technical claims. The broadcasts claimed, among other things, that voting machines switched votes and that an algorithm skewed results. Each such claim is falsifiable precisely because it is specific. “The system does X under conditions Y” is a sentence you can try to knock down. After testing it either stands or it doesn’t. (Rare among public arguments: this one has an answer key.)
The second is the internal messages. The record before the court included texts and emails from hosts, producers and executives at the network, gathered in the proceeding and quoted in the opinion. Dominion argued the messages showed what those people actually knew and believed, against what was aired. Whether that amounts to actual malice was reserved for the jury and never decided. But it is not a claim about software. It is material that devices and messaging systems created, and it can be tested too: who wrote it, when, from which device, and whether it changed since. (A dispute about what machines do, settled next to what devices quietly kept.)
You test the first to find out whether it is true. The second needs someone to show it is genuine. Good technical work separates what a system really does from a claim that sounds convincing but has no hold on it.
What a software expert witness actually tests when a technical claim is the case
Two different jobs, and both belong to the expert.
Job one: disproving a specific technical allegation about a system. You restate the allegation as behaviour you can test, then:
- Review the software and source code, where available, for the mechanism the allegation assumes.
- Read the audit logs and the audit trail the system leaves.
- Compare against the paper record, where there is one: paper is an independent check on the electronic count.
- Reproduce the alleged behaviour in a controlled environment and document whether it occurs, and if it cannot, why.
- Write it to survive cross-examination: each step reproducible by the other side, with no “trust me” anywhere.
Job two: authenticating and scoping internal-message discovery. A message is evidence only if you can show it is what it appears to be:
- Collect from the real custodians and devices, not from a copy someone forwarded.
- Preserve the metadata: timestamps, sender and recipient identifiers, message IDs.
- Reconcile the export against the source system, so the file produced matches what sits there.
- Scope by custodian, date range and keyword, not wholesale.
Same discipline both times: don’t trust the description, read the record. That is how a computer forensics expert witness works, whether the job is disproving a claim about a system or authenticating a message offered as evidence.
Why this matters beyond one case
There is no Israeli ruling here and I am not claiming one. But the structure is familiar to anyone litigating defamation or commercial disputes. In my own cases, more and more of these disputes turn on a claim about someone’s product or system. “Their system falsifies data.” “The algorithm discriminates.” “The product leaks information.” Each is a claim about software, and the rebuttal is a software opinion, not a speech.
Two practical points for the next file. First, restate the claim early as behaviour you can test; that tells you which test decides it. Second, treat the other side’s digital evidence (messages, emails, chats) as a collection problem from day one. A message gathered without chain of custody or metadata can be excluded or weakened once someone examines it.
The above is general information and not legal advice. Case facts are drawn from the sources listed.