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-132
Social media and online platforms · United States · 2020 · Slow removal of live self-harm, then cross-platform spread via recommendation

A veteran's death on Facebook Live was slow to be removed and then spread across TikTok, YouTube and Instagram for weeks

By Ellie Harris · Filed Livestreamed 31 August 2020

Alleged: Facebook (now Meta); reuploads on TikTok, YouTube, Instagram and Twitter 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.

A veteran's death on Facebook Live was slow to be removed and then spread across TikTok, YouTube and Instagram for weeks

What happened

It was reported that on 31 August 2020, Ronnie McNutt, a US Army veteran, died by suicide during a broadcast on Facebook Live. Reporting indicates Facebook took close to three hours to remove the original livestream despite user reports, and the company later said it was reviewing how it could have taken the stream down faster. Facebook also pointed to reduced moderation capacity during the COVID period, with fewer human reviewers available.

After the original came down, reporting indicates the footage did not stay contained, and clips spread to TikTok, YouTube, Instagram and Twitter. On TikTok the video reportedly surfaced on users’ For You pages through algorithmic recommendation, which put graphic content in front of people who had not gone looking for it. A known evasion technique made this harder to stop, because some clips began with harmless footage before cutting to the graphic moment, which slipped past automated detection tuned to the graphic frames.

It was reported that a friend and podcast co-host of McNutt’s, Josh Steen, started a campaign, #ReformForRonnie, that pressed the platforms to change how they handle this kind of material.

What an auditable version would have shown

An auditable version would show how long the live report queue actually took to reach a human during a period of reduced staffing, and what the automated systems scored on the edited reuploads that opened with innocuous footage. It would show whether the recommendation engine had any suppression rule for content under active review for self-harm, or whether engagement signals kept promoting it while moderation chased copies.

Where the gap was

There were two gaps. The first was removal speed on the original live broadcast, which was measured in hours when self-harm demands seconds. The second was that the same recommendation systems built to spread engaging content spread this too, and detection could be dodged by a simple edit to the opening seconds.

What governance should have looked like

A live stream of someone harming themselves cannot sit in an ordinary queue for close to three hours. It needs a fast lane to a real person, with a limit measured in minutes, and enough people on shift that a bad stretch for staffing does not turn minutes into hours. That is a verification gate. Two other things would have helped. The recommendation engine should stop pushing a video while it is under review for self-harm, instead of promoting it to strangers at the same moment moderators are trying to pull it down. And detection should recognise the clip itself, not just the graphic frames, so a harmless few seconds spliced onto the front does not walk it past the filter. A conduct record of how long removal took, and what was recommended while it was happening, would let anyone check the response instead of taking the company’s word for it.

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

This entry concerns suicide and self-harm. If you or someone you know needs support in Australia, Lifeline is available on 13 11 14 and the Suicide Call Back Service on 1300 659 467.

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 Facebook (now Meta); reuploads on TikTok, YouTube, Instagram and Twitter 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 →