Whistleblowers claim that TikTok and Meta jeopardized security in order to win the algorithm arms race.
Whistleblowers claim that TikTok and Meta jeopardized security in order to win the algorithm arms race.
Whistleblowers told the BBC that after internal research into their algorithms revealed how indignation fueled engagement, social media giants made decisions that enabled more harmful content on people's feeds.
As the firms fought for users' attention, more than a dozen insiders and whistleblowers exposed how they took chances with safety on matters like terrorism, sexual blackmail, and violence.
In order to compete with TikTok, a programmer at Meta, the company that owns Facebook and Instagram, explained how senior management had instructed him to permit more "borderline" damaging content, such as misogyny and conspiracy theories, in user feeds.The engineer stated, "They sort of informed us that it is because the stock price is down."
A TikTok employee provided the BBC with exclusive access to the company's internal dashboards of user complaints, along with additional proof that staff members had been told to give priority to a number of cases involving politicians over a number of reports of offensive posts showing minors.
According to the TikTok employee, decisions were made to "keep a strong relationship" with political officials in order to prevent threats of regulation or bans rather than because of the hazards to users.
The industry's reaction to the sudden expansion of TikTok, whose highly engaging algorithm for recommending short videos upended social media and left rivals scurrying to catch up, is shown in detail by the whistleblowers who spoke to the BBC documentary Inside the Rage Machine.
Matt Motyl, a senior Meta researcher, claimed that Instagram Reels, the company's rival to TikTok, was introduced in 2020 without enough security. According to internal research provided to the BBC, hate speech, violence or incitement, and bullying and harassment were much more common in Reels comments than elsewhere on Instagram.
According to Matt Motyl, Meta introduced Instagram Reels without enough security.
According to another former senior Meta employee, the firm invested in 700 employees to expand Reels, but safety teams were denied two specialists to handle child protection and ten more to assist with voting integrity.
Dozens of "high-level study documents revealing all sorts of hazards to users on these networks" were sent by Motyl to the BBC. Evidence that Facebook was aware of issues brought on by their algorithm was among them.
According to one internal research, the algorithm gave content producers a "route that maximizes revenues at the price of their audience's wellness" and the "current set of financial incentives our algorithms create does not appear to be aligned with our objective" to promote global unity.
Facebook may "choose to be idle and keep giving users fast-food, but that only works for so long," according to the statement.
Meta responded to the allegations made by the whistleblowers by saying, "Any idea that we purposely amplify bad content for commercial advantage is incorrect." According to TikTok, these were "fabricated accusations" and the business made technological investments to ensure that harmful information was never watched.
According to Ruofan Ding, a machine-learning expert who worked on TikTok's recommendation engine from 2020 to 2024, the algorithms are a "black box" whose internal workings are hard to examine.
He claimed that creating systems like this that were totally safe was challenging. "The deep-learning algorithm itself is beyond our control."
According to him, the developers do not give the posts' substance much thought. "To us, all the content is just an ID, a different number."
Comparing their relationship to various teams working on different pieces of an automobile, he said they relied on the content safety teams to make sure that dangerous posts were removed so they could not be boosted by the algorithm.There is the group in charge of the engine and acceleration, correct? Therefore, we anticipate that the team working on the brake system was performing well," he stated.
However, Ding claimed that while TikTok attempted to enhance its algorithm nearly every week in an effort to increase its market share, he began to see more "borderline" or problematic posts that only surfaced after people had been watching videos for some time.
Social media firms typically use the term "borderline material" to refer to posts that are detrimental but legal, such as conspiracy theories and those that are racist, misogynistic, or sexualized.
Teenagers told the BBC that major social media sites continue to recommend violent and hateful content despite the systems for users to express they do not want to see problematic content not functioning.
In one extreme instance, Calum, a 19-year-old, claimed he had been "radicalized by algorithm" since he was 14. He claimed that the algorithm displayed content that infuriated him and caused him to embrace racist and misogynistic beliefs.
Calum claimed that starting at the age of 14, he was "radicalized by algorithm."
He claimed that the videos "energized me, but not really in a good manner." "I was just a little upset with them. It accurately expressed how I felt on the inside, which was that I was upset with everybody around me."
In recent months, antisemitic, racist, violent, and far-right posts have become more common, according to UK counterterrorism police experts who examine thousands of messages on social media each.According to one cop, "people are not hesitant to express their opinions and are more desensitized to real-world violence."
"Delete TikTok"
Over the course of several months in 2025, the BBC had regular conversations with Nick, a member of TikTok's trust and safety team. On his laptop, we were able to see the company's internal dashboard, which described the instances his particular trust and safety team was handling and how it handled them.At some point, you can decide whether or not to speak up if you are always feeling bad about what you have been told to do," Nick remarked.
Teenagers and children were particularly at risk since it was too difficult to monitor the volume of cases they were evaluating to keep users safe, he continued. He believes that the company's capacity to handle this type of content has been hampered by budget cuts and the reorganization of some moderation teams, where AI technology is replacing some responsibilities.
According to the whistleblower, information about "terrorism, sexual assault, physical violence, abuse, trafficking" seems to be growing.
"Very different in a lot of ways to what the sites are stating" in public, he continued, is the reality of what the app advises and the action taken against hazardous information.
Nick demonstrated to the BBC how TikTok gave certain relatively unimportant issues involving politicians a higher priority for the safety team to analyze than a number of complaints regarding teenage danger.
In one instance, a 17-year-old who reported experiencing cyberbullying and impersonation in France and a 16-year-old in Iraq who complained that sexualized photos purporting to be of her were being shared on the app were given less priority than a political figure who had been made fun of by being compared to a chicken.
Some critical incidents involving children were identified by TikTok's internal systems as P2, or lesser priority, as seen in this BBC reconstruction.
"If you look at the place where this complaint originates from, it is really high risk because it is a juvenile and it involves sexual blackmail and then you can see the priority here," Nick stated, referring to the Iraq case. There is not much urgency.
Additionally, Nick provided instances where messages urging people to commit crimes or join terrorist organizations were not marked as top priority.
The whistleblower claimed that when the trust and safety team requested that cases involving young people be given priority over political cases, they were ordered not to and to keep handling the cases in accordance with the ranking they had been given.
According to Nick, the company prioritizes having a "strong relationship" with politicians and governments over children's safety in order to avoid laws or prohibitions that could negatively impact its business.
Because they "are not being exposed to this stuff on a day-to-day basis," Nick claimed that when he and other employees brought up some of these issues with management, they were unresponsive.
"Delete it, keep them as far away as possible from the app for as long as possible," the trust and safety employee bluntly advises parents whose kids use TikTok.
TikTok stated that it "fundamentally misrepresents the way their moderation systems operate" and refuted the notion that political content takes precedence above youth safety.
Nick's team is a part of a larger safety system that comprises several teams in charge of examining content complaints. "Specialist procedures for some concerns do not result in the deprioritization of child safety matters, which are addressed by dedicated teams within parallel review structures," TikTok stated.
The critique, according to a TikTok representative, "ignores the reality of how TikTok enables millions to discover new hobbies, find community, and supports a thriving creator economy."
According to the corporation, accounts for teenagers include over fifty pre-configured safety settings and features that are activated immediately. Additionally, it stated that it maintains stringent recommendation criteria, invests in technology that helps prevent hazardous information from ever being viewed, and offers options that allow users to customize their experiences.
"Do all in our power to catch up."
When Instagram introduced Reels in 2020 in reaction to TikTok going viral during the Covid epidemic, the algorithm arms race became more intense.
According to Matt Motyl, a senior researcher at Facebook and its 2021 successor company Meta, this was an attempt by the corporation to "copy" the "unique offering" that TikTok had introduced.
Between 2019 and 2023, he tested how material was prioritized in feeds by "conducting large-scale trials on sometimes as many as hundreds of millions of users" who frequently had "no idea" this was happening.More than three billion people use Meta's products, and the longer you stay on them, the more advertisements they sell, and the more money they make. However, it is crucial that they do this correctly because when they don't, terrible things occur," he stated.
Regarding Reels specifically, Motyl stated that the strategy was to proceed as fast as possible regardless of the effect on consumers. A "common trade-off between protecting people from hazardous content and engagement" exists, according to him.
According to one research paper he supplied with the BBC, Meta was having trouble preventing harm on Reels after its introduction. It indicates that negative comments were more common on Reels posts than on the main Instagram feed: 75% more for bullying and harassment, 19% more for hate speech, and 7% more for solicitation to violence.
He said that there was a "power imbalance" because teams in charge of Reels had to approve any new features or products that would increase user safety before safety personnel could implement them. According to Motyl, the Reels employees had "incentives to not let those goods debut since hazardous material generates more interaction than non-toxic."
During some senior-level conversations, Brandon Silverman—whose social media monitoring platform Crowdtangle was acquired by Facebook in 2016—described how CEO Mark Zuckerbeg was "extremely worried" about competition.No amount of money is too much when he perceives possible competitive forces, according to Silverman.
He claimed that although the corporation was concentrating on growing Reels during this time, safety teams were having difficulty obtaining permission to hire small numbers of employees. Another team said, "Oh, we just received 700 for Instagram Reels." He remarked, "I was like, OK yeah."
Mark Zuckerberg was "extremely concerned" about competition, according to Brandon Silverman.
According to Tim, a former Meta engineer, more potentially dangerous content was permitted on the platform as the firm tried to compete with TikTok. Until the "commercial positioning" shifted, his team had been concentrating on cutting this content.Your stock price must decline since you are losing to TikTok. People became reactive and suspicious, thinking, "Let us just do all we can to catch up." Tim asked, "Where can we obtain like 2%, 3% revenue for the next quarter?"
He claimed that a senior vice-president of Meta, who Tim assumed reported directly to Mark Zuckerberg, made the decision to stop restricting content that users were interacting with that was potentially dangerous but not illegal.
Internal records that Motyl, the senior researcher, released with the BBC at the period when Facebook was claiming to be merely a "mirror to society" show how the firm was aware that it was promoting content that incited anger and even violence.
The documents describe how sensitive information, such as messages that instigate violence or touch on people's moral convictions, is more likely to elicit a response and interaction on the website, particularly if it incites outrage.According to the report, "our algorithms assume that people appreciate the material and want more of it given the disproportionate involvement."
According to Silverman, there was a point when Meta was "genuinely introspective" and the company's leadership originally appeared unclear of how to handle harmful information on the site.
However, he claimed that they "began to calcify into a sort of defensiveness" in their position. He stated, "We are not accountable for all the polarization in society."No one is accusing you of being the cause of all polarization. All we are saying is that you contribute to it, most likely in ways that you do not have to. You might not contribute as much if you only made a few adjustments," Silverman remarked.
The whistleblowers' allegations were refuted by a Meta representative. The representative stated, "The truth is, we have rigorous standards to protect users on our platforms and have made considerable investments in safety and security over the last decade."
With the launch of a new Teen Accounts feature "with built-in protections and tools for parents to manage their adolescents' experiences," the business claimed to have "made meaningful changes to protect teens online."
