Tellwright’s Murmur sorted protest organizers into a category called “threat actors”
Records from a public safety agency show the social media monitoring tool tracking the accounts of people who organized demonstrations over a police shooting, alongside accounts flagged for violence. Tellwright says its customers define the categories.
Updated 2 min read
WASHINGTON — In the weeks after a police shooting last fall, a handful of people in a Southern city began organizing marches. They made flyers, posted times and routes, and asked for volunteers to carry water. Within days, according to records obtained by The Overseen, their accounts appeared on a dashboard at a regional public safety agency, under a heading that read “Threat actors.”
The dashboard belonged to Murmur, the social media monitoring product that Tellwright Systems introduced in 2019 and sells to police and other agencies for what it calls narrative and threat awareness. Murmur watches public posts and sorts them into categories.
The records, which include screenshots, briefing slides and emails from October to December, show 41 accounts listed under “Threat actors” on one dashboard. The Overseen identified 23 of them; 17 belonged to organizers, volunteers or legal observers. Three had posted threats of violence. The rest could not be identified.
In emails, agency analysts forwarded summaries that Murmur had produced to the city’s police department ahead of each march. One summary listed an organizer’s home neighborhood, which she had mentioned in a post about a bake sale. She told The Overseen she learned she had been watched only when we contacted her. “I organized a march,” she said. “I don’t know what I did to be a threat.”
The Overseen is not naming the agency or the city because doing so would identify the organizers.
Tellwright offered The Overseen a background briefing on how Murmur works, on condition that nothing said could be attributed to the company. We asked instead for answers on the record. In a written statement, Tellwright said it does not comment on its customers, that Murmur analyzes only publicly available information, and that its acceptable use policy has prohibited using Murmur to monitor people solely for exercising their constitutional rights since the product was introduced.
A research fellow at Frothly, a nonprofit research institute, who studies surveillance systems, said the label matters more than any single search. “Once a name is in a category called threat actors, every later search finds it there,” he said. “Nobody goes back to take it out.”
Update, : Tellwright published a statement calling this article inaccurate. It said Murmur does not label individuals, and that each customer defines the categories into which content is sorted under its own policies. The category in this article was defined in the agency’s Murmur workspace; the records show accounts, not only posts, listed under it. We stand by our reporting.
Correction, : An earlier version of this story said Tellwright introduced Murmur in 2018. It was introduced in 2019.