Photos by Michael Starghill
On the afternoon of June 1, Kaitlyn Siragusa received a notification from Instagram: One of her posts had been taken down for violating the platform’s content policies. If she were to break the rules again, the alert warned, her 1.6 million-follower account could be disabled. Siragusa, a 25-year-old Twitch streamer known to fans as “Amouranth,” was puzzled. She was confident that the photo, which showed her standing poolside in a striped bikini, fully complied with Instagram’s policies.
It was the latest of nearly a dozen posts that Instagram had recently deleted, including some that she had uploaded long before their removals. In one case, she got a notification alerting her to the deletion of an “Instagram Story” that had publicly expired from her profile 10 days earlier. (Stories automatically disappear after 24 hours, though they may be saved to the poster’s private archives.) The streamer’s content had been disappearing so frequently over the previous several weeks, Siragusa told HuffPost, that she’d started a spreadsheet to keep track.
Four days later, on June 5, Siragusa received a mysterious email from someone claiming to work for Cognizant, an American tech firm hired to do content moderation for Facebook and Instagram.
“I’m sure you’ve noticed recently than [sic] many of your posts and stories have been removed,” said the email, one of several that Siragusa shared with HuffPost. “Perhaps we can reach an agreement privately.”
For monthly payments of 0.25 bitcoin (about $2,600), the emailer proposed, Siragusa could rest assured that her content would stay up. The emailer signed off as “Tampa” — the location of a Cognizant facility where content moderators have described being severely overworked and underpaid.
At first, Siragusa ignored the email. She’d never heard of Cognizant, and she assumed she was just being targeted by a scammer who was reporting her posts as violating Instagram’s rules over and over again until they disappeared and then posing as a content moderator to try to extort her. But there was something odd about the next messages that “Tampa” sent her in July, following more photo takedowns.
“I am sure by now you have noticed that two additional posts have been struck and removed. Your posts made on April 3 at 10:18 AM, and July 2 at 11:54 AM,” said one of the emails. Another simply stated: “July 12 at 10:34 AM.”
When a photo or video is posted to Instagram, it doesn’t feature a publicly visible timestamp, just a date. There are little-known ways to manually extract timestamps from existing posts — but not for those that have already come down. And yet, somehow, “Tampa” had correctly listed the exact upload times of Siragusa’s recently deleted Instagram photos, including one that she’d posted more than three months earlier.
Siragusa was able to confirm the times by checking the removal notifications she’d received from Instagram, such as the one shown above on the left. (After taking down a user’s content, Instagram sends them a private notification including a blurred copy of the post, the reason for its deletion, the date and time it was posted, and a link that can be used to appeal the removal.)
“When I saw the timestamps, that’s when I got concerned,” said Siragusa, who earns a living by video-streaming and online modeling, and relies on Instagram to promote herself. “I realized this was probably an actual person with Instagram.”
Siragusa had her agent get in contact with Facebook, which owns Instagram, to discuss the matter last month. The agent was able to arrange and record a phone call with a Facebook representative ― a rare line of communication that’s normally unavailable to the vast majority of Facebook users. Audio of the recording was shared with HuffPost.
During the 20-minute call, Siragusa’s agent explained three times that a self-proclaimed Instagram content moderator was trying to extort Siragusa for content protection.
“I can’t answer for the emails that you’re receiving,” responded the representative, who adamantly repeated that Siragusa’s posts had been “rightfully removed” for being “sexually suggestive.” The poolside photo was deemed inappropriate for Instagram in part because Siragusa had “her hand near her chest, and things like that,” according to the representative. A photo of her in a Hooters jersey, which was flagged as “nudity or pornography,” was removed because “she’s, uh, pouring water on herself.”
In the spring of this year, Instagram quietly started shadow-banning users’ borderline content that does not actually violate the company’s community guidelines, including vaguely “inappropriate” and “sexually suggestive” posts. Such content is subject to algorithmic demotion but, per Instagram’s own public policies, not deletion. When Siragusa’s agent reminded the Facebook representative that “sexually suggestive” content does not meet Instagram’s criteria for removal, the representative doubled down, stating: “We can’t allow that kind of content on our platform.”
Frustrated and confused, having received no real answers from Facebook despite her rare direct access, Siragusa reached out to HuffPost.
“I feel like it is almost impossible for individuals — even those with millions of followers — to face down Facebook,” she said.
Contacted by HuffPost, Cognizant declined to comment for this story, instead deferring to its client. In a statement, a Facebook spokesperson said that the representative with whom Siragusa’s agent had spoken is part of a team that “isn’t versed in the details of the Community Guidelines nor is it their responsibility to communicate them.”
Regarding the emails Siragusa received from “Tampa,” Facebook told HuffPost: “We take accusations like this very seriously. We investigated this matter and did not find any evidence of abuse.”
According to the Facebook spokesperson, Instagram also determined that all three of the deleted posts for which “Tampa” provided timestamps had been flagged to content moderators by Instagram’s proactive artificial intelligence technology — not from user reports — as spam, nudity and sexual solicitation.
The fact that Siragusa’s posts were flagged by AI doesn’t wholly eliminate the possibility that “Tampa” is just an opportunistic scammer out in the world, with no relation to Cognizant or Instagram. But it does make that possibility rather unlikely. With AI flagging the content, “Tampa” would have had to diligently track each of Siragusa’s Instagram pictures, manually collect the timestamps for all of them, wait for them to go down on their own, and then contact her with the posts’ details.
This extremely patient outside scammer — who apparently never bothered to speed things up by reporting Siragusa’s posts — must have also known about Cognizant’s Tampa facility at the time of the initial email to Siragusa, which was sent weeks before the first press mention about Cognizant’s content moderation operations in the city.
“Tampa” also exhibited precise knowledge of non-public Instagram policy. In that initial email, while trying to entice Siragusa to pay for protection, “Tampa” told her that content rules can be “strictly or loosely enforced” and that once Instagram removes a user’s post, the user’s account is restricted for exactly 14 days, during which impressions generated by the account are limited.
Instagram confirmed to HuffPost that the information about the 14-day restriction period is accurate and that the company had not shared it publicly.
“Tampa” did not respond to HuffPost when contacted by email. It is unclear if he or she has also solicited bribes from other Instagram users.
Over the past several days, three more of Siragusa’s Instagram photos have come down, right at the end of her most recent 14-day restriction period — meaning her account’s engagement will be limited for another two weeks at least. This time, the accompanying Instagram notifications didn’t give specific reasons for the removals — stating only that each post “goes against Community Guidelines” — and offered no option to appeal.
It’s disconcerting that Instagram seems to “pick winners and losers by fiat,” said Siragusa. “There is something extremely suspect in the way this whole ordeal has played out.”
Have you been directly contacted by a self-proclaimed content moderator? Let us know by emailing jesselyn.cook@huffpost.com or scoops@huffpost.com.