180 incidents on record · 2026 Headlights Incident reports by Ellie Harris · Melbourne
10 new this week Library last updated 30 August 2026
← The incident library
HD-INC-138
Social media and online platforms · Slovakia · 2023 · Manipulated audio spread during an election, with no clear platform response

Two days before Slovakia's election, a faked audio of a party leader discussing vote-rigging spread during the campaign silence

By Ellie Harris · Filed Audio spread around 28 September 2023; election 30 September 2023

Alleged: Meta (Facebook) as the main platform; the source of the audio was not publicly established developed or deployed the AI system implicated in this incident. Details are drawn from public reports; parties are presumed innocent of any wrongdoing not established by an official finding.

Two days before Slovakia's election, a faked audio of a party leader discussing vote-rigging spread during the campaign silence

What happened

In the final days before Slovakia’s 2023 parliamentary election, an audio clip spread across social media that appeared to capture the leader of the Progressive Slovakia party and a journalist discussing how to rig the election, including buying votes.

The people heard in the recording said it was fake, and fact-checkers concluded the audio showed signs of being generated or manipulated using AI. The conversation never happened.

The timing made the situation worse. The clip appeared during Slovakia’s 48-hour election silence period, when campaigning and media coverage are restricted. That made it much harder for those targeted to respond quickly or widely. Fact-checkers and the platforms did react, but reports at the time suggested many of the platform’s manipulated media rules were written mainly for video, leaving AI-generated audio in a grey area. Meta later expanded its approach to include AI-generated audio, images and video, but by then the election had already taken place. It isn’t possible to know what impact, if any, the clip had on the result.

What an auditable version would have shown

An auditable system would have shown exactly what happened behind the scenes. It would record when the audio was first reported, who reviewed it, how quickly it was assessed, what action was taken, and whether its timing during the election silence period triggered a faster response. Instead of guessing what the platform knew and when, there would be a clear record of the decisions that were made.

Where the gap was

There were two obvious gaps. The first was format. The systems designed to detect manipulated media were largely built around video, while this incident involved audio. The second was timing. A suspected AI fake appeared during the short period before voting when those targeted had the least opportunity to respond. There was no clear fast-track process recognising that an election-related deepfake in the final hours before polling carries a much higher risk than ordinary misinformation.

What governance should have looked like

During an election, suspected AI-generated political content should be treated as high priority, regardless of whether it is audio, video or an image. A report like this should go straight to a specialist review team, with temporary labels or limits applied while it is being assessed. Each step should be recorded: when it was reported, who reviewed it, what evidence they considered, what action they took, and when. That doesn’t decide the election. It simply reduces the chance that a convincing fake gets hours of unchecked distribution at the moment it can cause the most harm.

A VerificationGate is designed to route a plausible political fake to a human quickly and to hold a label or a limit on it while the check runs, rather than letting it spread untouched. A ConductRecord of when the clip was flagged, what was decided, and what the audience was shown lets anyone judge the response afterwards.

The reference implementation of VerificationGate and ConductRecord is open source. It lives at github.com/saffronandindia/headlights-oss, Apache 2.0 licensed, free for any company to install. The repository is public now.

Sources

The mailing list

Fresh incident reports every week. One email to match.

We add new incidents to the library regularly, and send a single short email each week with what's new. The library stays free and open; this is just how you keep up with it.

No tracking. Unsubscribe in one click.

The record

An auditable system would have produced a signed, tamper-evident record the moment this happened: what the system did, the version that did it, the basis it acted on, and the action taken, and Meta (Facebook) as the main platform; the source of the audio was not publicly established could have produced it on demand.

This is the record the system as deployed did not produce in a signed, auditable form.

What this teaches
Capture what happened when it happens
What the system did, the version that did it, the basis it acted on, and the action taken, recorded at the moment, not reconstructed after.
Sign it, so no one has to trust the record-keeper
A tamper-evident entry. Edit it later and the signature breaks. The record does not ask for the benefit of the doubt.
Make it verifiable by anyone
A court, a regulator, a customer's lawyer can check the record themselves, without taking the company, or us, at our word.

Headlights summarises publicly reported AI incidents. All summaries are independently written, attributed to their original sources, and intended for research and educational purposes. Allegations are identified as such until established through official findings.

This report is based on the sources listed above and reflects information available at the time of review; later developments may not be captured. Where a person is described as charged with or alleged to have done something, that allegation is unproven unless a conviction or a court or regulatory finding is stated. Headlights publishes journalism and commentary, not legal advice.

Want to write back?

Direct to my inbox.

ellie@useheadlights.com →