Send. Unsent. Suspended.
How political actors abuse Meta’s safety system to silence voices
A stranger sends you a message. Before you can read it a second time, it disappears. Minutes later, your Facebook account is suspended for allegedly violating Meta’s policies on human exploitation or adult sexual solicitation, without any evidence or explanation. Dismislab found that politically motivated actors are exploiting a possible vulnerability in Meta’s enforcement system to target journalists, activists in Bangladesh.
Zayan Ahmed Rony, a Chattogram-based activist, received a message on Facebook on May 21. Before he could read it, the sender unsent the message and then repeated the pattern several times. Thinking little of it, Rony ignored the messages.
Soon afterward, he received an email from Meta saying his Facebook account had been suspended for violating its policy on “adult sexual solicitation.” Although he briefly regained access while trying to recover the account, it was suspended again. He has been unable to restore it since.
Rony’s case was one of five examined by Dismislab. All five journalists and activists, whose accounts were suspended between May 21 and 22, said their Facebook accounts were suspended shortly after unknown accounts sent and unsent suspicious messages on Messenger. Their accounts of what happened followed a strikingly similar pattern.
The incidents came to the fore as a Facebook network calling itself “Crack Platoon Bangladesh Cyberforce” publicly claimed responsibility for a broader campaign. In a Facebook post on May 25, the group said it had in three days belonging to people it accused of “spewing filth centering Bangabandhu, Liberation War and Bangladesh.” Dismislab could not independently verify the claim. The network has previously been linked to coordinated campaigns targeting journalists, critics and activists, including a review-bombing operation documented by Dismislab.
The five cases examined by Dismislab suggest that politically motivated actors may be exploiting Messenger and Meta’s enforcement system to trigger account suspensions. By sending messages that Meta classifies as “human exploitation” or “adult sexual solicitation,” they appear to be creating situations that can lead to the temporary or permanent suspension of journalists’ and activists’ Facebook accounts.
The technique is so widespread that Dismislab identified about a dozen online tutorial videos that claim to show users how to suspend a Facebook account within minutes by exploiting Messenger and Meta’s enforcement policies.
Dismislab asked Meta whether it was aware of the apparent abuse of Messenger, how it detects attempts to exploit its enforcement systems, and whether it plans to address the issue. The company did not respond.
The victims interviewed by Dismislab said losing access to their Facebook accounts cut them off from an important channel for reporting, advocacy and communication. According to them, attack as such can have a chilling effect on expression.
Five accounts, one pattern
All five people interviewed by Dismislab described a strikingly similar sequence of events. The process begins with contact from an unfamiliar Facebook profile, sometimes in the form of comments on posts, other times a direct message through Messenger. The ultimate goal is to make the victim “read” the message. Once the victim opens the message, the unfamiliar account sends a long block of text, then quickly unsends it. They often repeat the sending and unsending process multiple times. Before the recipient has time to understand what happened, they discover that their Facebook account has been suspended. Below is what happened to each of the five victims. Scroll to move through each case.
Zayan Ahmed Rony
Rony received a message in his inbox on May 21, which was immediately deleted by the sender. The sending and unsending of messages took place multiple times. Rony read enough to understand that it contained something sexual, and didn’t reply. Soon after, he received an email from Meta stating his verified account was suspended for adult sexual solicitation. He got the account back for a short time. Then he lost it again, for good — along with years of memories.
Qadaruddin Shishir
A stranger kept asking for help on multiple of his posts. Then on May 21, the same profile sent a message to Shishir on Messenger, but when he tried replying to it, his profile vanished mid-conversation. Among the messages he never finished reading, one had been sent, then unsent, before he could see it. Meta cited “human exploitation” as a reason for suspending his account but did not mention the specific incident.
Zyma Islam
A profile with a woman’s name asked for help through Messenger on May 22. The same profile then called her, but Zyma did not answer. However, soon after this incident, her account was suspended. She was targeted once again by the same profile on June 30. Shortly after receiving a message from the same account, Meta suspended her profile again. However, she recovered her account on July 10.
Md Saifullah
An account with a female name sent him some messages or long texts resembling codes on May 22. These messages were sent and then unsent. When this was repeated multiple times, he did not reply. After a short while, he discovered that his account had been suspended. He also said he knew nothing about why it was suspended.
Kazi Majharul Islam
He received a message seeking help on May 22. Then a long message was sent and deleted repeatedly. The profile then asked him to check messages sent from another account. The second profile repeated the same thing. Within two or three minutes, his account was suspended. He appealed to Meta, but has yet to get the account back.
How reporting is carried out
All five victims told Dismislab that their accounts were suspended shortly after receiving messages from unfamiliar Facebook profiles.
To better understand how these targeted attacks are carried out, Dismislab reviewed publicly available materials describing methods of reporting accounts and triggering enforcement actions. Searches on Facebook led to several videos and posts demonstrating a process that closely resembled the experiences shared by the victims.
The videos appeared on pages and groups with names such as Sabbir Cyberlab, Bangladesh Cyber Squad, Shadow Core Recovery Unit, and RDCS Cyber. They described a sequence of sending messages containing language related to human exploitation or sexual solicitation, removing the messages, and then reporting the conversation to Meta.
Similar instructions were also found outside Facebook. Dismislab identified a post in multiple Telegram groups explaining account reporting techniques, a YouTube video demonstrating a similar process and a publicly accessible note on the web utility platform Voz. The YouTube video’s Vietnamese title translates as “Latest message reporting/banning tutorial,” while its description contained a text related to sexual transactions written in a mixture of English and Filipino.
The presence of similar instructions across multiple platforms and languages suggests that the method may extend beyond Bangladesh and could be part of a broader tactic used to exploit Meta’s enforcement systems.
Five moves, one suspended account — reconstructed from victims’ testimony and online tutorials.
Approach
An unfamiliar profile initiates contact — through a comment, a follow or a friend request — often asking for help or trying to start a conversation.
Approach
An unfamiliar profile initiates contact — through a comment, a follow or a friend request — often asking for help or trying to start a conversation.
Message request
The profile sends a message request, creating a conversation thread that may later become part of the reporting process.
Solicit
A message containing language related to human exploitation or sexual solicitation appears in the conversation — the type of content prohibited under Meta’s policies.
Sent, unsent
The message is quickly unsent, sent and unsent again multiple times before the recipient can capture evidence. In some cases, a second profile repeats the sequence.
Suspended
Minutes later, a notice arrives. The account is suspended. Victims say they are never notified which specific message or action triggered Meta’s decision.
Meta Policy, AI enforcement and unanswered questions
The messages used to target the victims appear to intersect with two of Meta’s Community Standards: adult sexual solicitation and human exploitation. Meta’s Adult Sexual Solicitation and Sexually Explicit Language policy states, “We also restrict sexually explicit language that may lead to sexual solicitation because some audiences within our global community may be sensitive to this type of content.” Under its Human Exploitation policy, the company says, “We may remove accounts based on their interactions with other violating accounts, searches for or interactions with violating content, or membership in communities (e.g. Groups) we have removed for violating our policies.”
The messages used to report the victims’ accounts also appear to violate these policies. , for example, contained language offering sexual services. Among the five victims interviewed by Dismislab, three said Meta cited adult sexual solicitation as the reason for suspending their accounts.
To identify policy violations, Meta expanded its use of AI-powered enforcement in March this year. Announcing the rollout, Meta explained that its AI systems “catch two times more violating adult sexual solicitation content than our review teams, while also decreasing the rate of mistakes by more than 60%.”
Dismislab could not independently determine whether the victims’ accounts were suspended specifically because the policy-violating messages were sent and then reported, or whether AI-driven moderation played a role in the enforcement decisions. However, the messages received by the victims, the reasons Meta cited for the suspensions, and the reporting process demonstrated in publicly available tutorials indicate violations of Meta’s policies and the involvement of an automated enforcement system.
A recent incident in India has also confirmed that Meta is actively using AI-driven automated enforcement. Meta removed a Facebook post by Indian Prime Minister Narendra Modi. They later informed the Indian government that the removal had resulted from an AI-powered automated content moderation system.
Dismislab contacted Meta to ask whether a vulnerability in its enforcement system could be exploited to take down journalists’ and activists’ accounts through Messenger reporting, and, if so, how the company plans to address the issue. Meta did not respond to the request for comment.
Network behind the takedown campaign
Dismislab investigated who was behind the takedown campaign targeting journalists and activists by exploiting an apparent vulnerability in Meta’s enforcement system. One name repeatedly emerged: Crack Platoon Bangladesh.
The origins of the network trace back to the aftermath of Bangladesh’s July-August 2024 mass uprising, which ended the Awami League’s 17-year rule. On October 11, 2024, supporters of the ousted party created a Facebook page called Crack Platoon Bangladesh to promote the Awami League and criticize its political opponents. Since then, the page has regularly published content supporting the party and criticizing rival political groups.
On March 3, 2025, another Facebook page, Crack Platoon-Bangladesh Cyberforce, was created. Since its creation, the page has repeatedly claimed responsibility for disabling Facebook profiles and pages belonging to political opponents. It has also celebrated the restoration of accounts belonging to users it identifies as supporters.
Dismislab previously found that several accounts associated with a private group with the same name, Crack Platoon-Bangladesh Cyberforce, participated in a coordinated review-bombing campaign. The operation sought to damage the reputations of Facebook pages belonging to news organizations, political opponents and critics by coordinating large numbers of negative reviews.
Several of the accounts that promoted the review-bombing campaign also appear to be connected to the Crack Platoon Bangladesh network involved in the recent account takedown campaign. The network map below illustrates how those pages, profiles and groups are connected.
Cost of losing an account
When someone’s Facebook account is suspended, it has both personal and professional consequences. It means more than just losing access to Facebook. Many of the victims lost years of photos, videos, and other content documenting significant events that had not been backed up elsewhere. Others lost a follower base they spent years building, meaning their voices would not reach the intended audience.
The suspensions also disrupted journalistic activities, cutting off an important channel for reporting and public engagement. Those affected said the experience left them feeling that this takedown campaign could be used to silence journalists and independent voices, raising broader concerns about digital censorship and the vulnerability of online platforms. They also expressed apprehension about political opinions being muffled, which would curtail freedom of democratic expression.
I posted a lot of videos and photos during the July movement. I can’t get them back.
When I wrote something, people read it. I don’t have the time to build that audience again.
It seriously hampered my work as a journalist.
If this can happen to me, it can happen to anyone viewed as a political opponent.
Beyond suspended accounts
The challenges rarely end with the suspension itself. A report by New York-based Richt Law Firm found that users frequently struggle to recover disabled Facebook accounts through Meta’s largely automated appeals process. The report describes financial losses for businesses, reputational harm stemming from severe policy violations, and emotional distress as users encounter repeated automated rejections with little explanation.
TechCrunch’s reporting on Meta’s Oversight Board lays out just how threadbare the company’s response to these bans really is. The board’s review began with a single case: a journalist was threatened with violence. While reviewing it, the board discovered what TechCrunch describes as broader “systemic human rights concerns” related to Meta’s two-track system for punishing accounts, one for accumulating strikes and another for permanently disabling accounts for “egregious” violations, with no clear or documented line separating the two.
The board also found that paying for Meta Verified does not necessarily result in meaningful support when an account is disabled. Although the service promises enhanced customer assistance, the board concluded that verified users often faced the same opaque appeals process as everyone else. One of the people interviewed by Dismislab, Zayan Ahmed Rony, had a verified Facebook account when it was suspended.
The Oversight Board has urged Meta to provide users with a record of their policy violations, clearer explanations of enforcement decisions and greater transparency about the role of automated systems in account moderation. Until those safeguards are in place, cases like those documented by Dismislab raise broader questions about whether coordinated actors can continue exploiting Meta’s enforcement mechanisms to silence critics while leaving victims with little explanation and limited avenues for appeal.