Courtney's House is located on a quiet residential street just outside the capital. Frundt and her team have tried to make the modest two-story house feel like a family home, with comfortable couches and photos on the mantel. Frundt, who was a victim of child trafficking in the 1980s and 1990s, is now one of the most experienced and respected advocates in the nation's capital. Caring and fiercely protective of the minors in her care, she is hired by the city's child protective services to identify trafficked children who come through the courts, and she regularly attends court hearings involving youths residing at the shelter. She also helps train the FBI and local law enforcement sex trafficking units on how to spot traffickers on online platforms such as Instagram. “A long time ago, when I was a victim of trafficking, I was advertised in the classified sections of free newspapers,” Frundt tells us. “Now the young people we work with have been trafficked on Instagram. It’s exactly the same business model, only you don’t have to pay to place an ad anymore.”

What makes social media platforms such a powerful tool for traffickers — far more powerful than the back pages of a newspaper where Frundt advertised as a teenager — is the way they allow them to identify and cultivate relationships with both victims and potential clients. Traffickers can post ads and negotiate through different features of the same platform: sellers sometimes post information about the girls they have available, then switch to direct private messages to agree on a price and provide the meeting location to the client.

US court documents clearly show how these platforms are used. In one case tried in Arizona in 2019, Mauro Veliz, a 31-year-old man who was convicted of conspiracy to commit sex trafficking of a minor, exchanged messages on Facebook Messenger with Miesha Tolliver, who was also sentenced to prison for sex trafficking. Tolliver told Veliz he could provide him with a girl to have sex with, and showed him photographs of two more, before saying the girls were 17, 16 and 14 years old.

—Veliz: “How much does it cost to have them all?”

—Tolliver: “The 14-year-old will be the most expensive… about $200. The others, about $150.”

Tolliver told Veliz that the 14-year-old was “new to sex.”

—Tolliver: “The one on the right… is 16 with a fat ass… the other one [is] 15 with huge tits.”

Court transcripts state that Veliz received several explicit sexual images from the girls.

—Tolliver: “Do you want me to bring one of the girls with me so you can fuck?”

[ ... ]

—Veliz: “Is your girl nervous? Or have you already told her?”

—Tolliver: “… she’s still young and doesn’t understand how people like her.”

Tolliver and Veliz exchanged more messages, agreeing that Veliz would meet the girl at a California hotel two days later. The last message presented to the court was from Veliz to Tolliver. “We’re done, he’s in the bathroom,” it said.

Luke Goldworm, a former assistant district attorney in Boston, Massachusetts, who has investigated and prosecuted trafficking cases for years, says he has encountered numerous exchanges like this. He notes that from 2019 until he left the job in October 2022, the number of cases of child trafficking crimes on social media that his department investigated increased at about a 30% annual rate. “We are seeing more and more people with criminal records moving into this area. It’s incredibly lucrative,” he says. A trafficker can earn up to $1,000 a night. Many of the victims in the cases he investigated were just 11 or 12 years old, and most were black, Latino or LGBTQI+.

According to Goldworm, although his investigations affected all social media platforms, Meta’s platforms are the ones he encountered most frequently. Six other prosecutors from different states told us that, in their experience, Facebook and Instagram are widely used to recruit and traffic children. Five of these prosecutors shared outrage at what they see as unnecessary delays by Meta in complying with court-signed orders and subpoenas required to gather evidence in sex trafficking cases. “We receive more rejected court orders from Facebook than from any other electronic service provider,” said Gary Ernsdorff, a senior deputy prosecutor in King County, Washington. “What I find frustrating is that the exchange can delay a victim’s rescue by a month.”

Three of these prosecutors described experiences in which they said the company claimed technicalities, pointed out wording and formatting errors, and slowed down investigations. In response, the company called these claims “false,” adding that between January and June of last year it “provided data in nearly 88 percent of U.S. government requests.”

Meta is aware that human traffickers use its platforms, but insists it does everything in its power to prevent this. By law, the company is required to report any images of child sexual abuse shared on its platforms to the National Center for Missing and Exploited Children (NCMEC), which receives federal funding to act as a national clearinghouse for child abuse information. Meta is a major funder of NCMEC and sits on its board of directors.

From January to September 2022, Facebook reported more than 73.3 million pieces of “child nudity and physical abuse” and “child sexual exploitation” content, and Instagram reported 6.1 million. “Meta leads the industry in using the most sophisticated technology to detect both known and unknown child exploitation content,” a company spokesperson said. Of the 34 million child sexual exploitation content removed from Facebook and Instagram in the last three months of 2022, 98% were detected directly by Meta.

From January to September 2022, Facebook reported more than 73.3 million pieces of content about "child nudity and physical abuse" and "child sexual exploitation," and Instagram reported 6.1 million

However, the vast majority of content Meta reports is child sexual abuse material (CSAM)—which includes pornographic photos and videos—rather than sexual exploitation. Unlike with child sexual abuse images, there is no legal obligation to report child sex trafficking, so NCMEC must rely on all social media companies to be proactive in seeking and reporting it. According to Staca Shehan, vice president of NCMEC’s analytical services division, this legal inconsistency—the fact that child sexual abuse images must be reported, but child sex trafficking is not required to be reported—is a serious problem.

“It’s concerning how underreported trafficking is,” says Shehan. Social media managers “prioritize what is [legally] required.” “I think all parties could do more,” says Shehan. “The volume of child sexual abuse material (CSAM) and the volume of trafficking [being reported] are light years apart.” According to Shehan, another reason for this disparity, beyond the different legal requirements, is technological. “Child sexual abuse material is much easier to detect. Many technological tools have been developed that allow this crime to be detected automatically.”

An NCMEC spokesperson told the Guardian that if big tech companies fail to report child sex trafficking, the crime is allowed to thrive online. They say reporting trafficking is crucial to saving victims and punishing offenders. According to records obtained through a subpoena seen by the Guardian, between 2009 and 2019, Meta reported only three cases of alleged child sex trafficking in the US to NCMEC. An NCMEC spokesperson confirmed this figure, but clarified that several cases of child trafficking during the same time period were reported by Meta under other “incident types,” such as child pornography or enticement. “I think one of the things to keep in mind is that there is a certain protocol used to report these situations,” Antigone Davis, Meta’s global security officer, stressed in a recent interview. “And just because something is not classified as sex trafficking doesn’t mean it’s not being reported.”

A Meta spokesperson said that over the past decade, the company had reported to NCMEC “tens of thousands of accounts that violated our policies against child sex trafficking and commercial child sexual abuse material.” When we raised these claims with NCMEC, it said it had not received “tens of thousands” of child trafficking complaints from Meta, but rather that it had received that number related to child abuse images.

Hany Farid is a professor at the University of California, Berkeley, who helped invent the PhotoDNA technology that Meta uses to identify criminal content. He thinks Meta, now valued at more than $500 billion, could do more to fight child trafficking. For example, it could invest more in developing better tools to “flag suspicious words and phrases in unencrypted parts of the platform, including coded language around child grooming.” “This is not fundamentally a technology issue, but rather a corporate priority issue.” (There is a separate debate over how to handle encryption. Meta’s plans to encrypt direct messages on Facebook Messenger and Instagram have recently drawn criticism from several agencies, including the FBI and Interpol.)

In response to Farid’s claims and further questions from the Guardian, Meta did not specify how much money it has invested in technologies to detect child sex trafficking, but said it had “focused on using AI and machine learning on the non-private, non-encrypted parts of its platforms to identify criminal content and accounts and make it easier for people to report messages to the company so we can take action, including referral to law enforcement.” Davis also stressed that Meta is constantly working with its partners to improve its safeguards against human trafficking. For example, she mentioned that “we have been able to identify the type of searches that users make when they want content linked to child sex trafficking, so that when users make this search, we will come up with information to divert them or let them know that what they are doing is an illegal activity.”

These efforts have not satisfied some of Meta’s own investors. In March, several pension and investment funds that own Meta shares filed legal action against the company in Delaware for its alleged failure to act in response to “systemic evidence” that its platforms are facilitating sex trafficking and child sexual exploitation. The lawsuit says the board of trustees has failed to protect the company’s interests by providing insufficient explanations for how it is addressing these crimes. Meta has rejected the basis of the lawsuit. “Our goal is to prevent people who seek to exploit others from using our platform,” the company said.

Several funds that have invested in Meta have taken legal action against the company in Delaware for its alleged failure to act in the face of "systemic evidence" that its platforms facilitate child sexual exploitation.

Meta relies on computer software but also human teams to moderate and identify cases of child grooming and sex trafficking. Until recently, Anna Walker* worked the night shift in an office for a Meta subcontractor. She started each shift feeling scared. “They would put us in a dark room to look at the material,” she says. Walker’s job was to review interactions between adults and children on Facebook Messenger and Instagram direct messenger that had been flagged as suspicious by Meta’s artificial intelligence software. Walker says she and her team struggled to keep up because of the workload. She says she saw cases of adults seducing children and then planning to meet them for sex, as well as conversations about paying for sex. Walker’s managers would pass these cases on to Meta to decide whether to take action against the user. In some cases, Walker says, “Months would go by and then the automated bot would send me an email saying it was closing this case, because no one had taken action on it.” She adds: “I would go to my boss to express my dismay at the cases I had raised and tell him that I wanted to help. I felt like no one was paying attention to these horrible situations.”

We spoke to six other moderators who worked for companies subcontracted by Meta between 2016 and 2022. Their account matches Walker’s. Their efforts to raise cases of possible child trafficking often fell on deaf ears. “In one post I reviewed, there was a picture of a girl who looked like she was about 12, wearing the tiniest lingerie you could imagine,” one former moderator explained. “It explicitly listed prices for different things, like the price of a blowjob. It was obvious that this was trafficking,” she said. She says her supervisor later told her that no action had been taken in this case.

When we relayed these accounts to Meta, a spokesperson said that moderators like Walker are not typically informed about whether the content they have reported is being investigated. They stressed that if a moderator doesn’t hear back about a case, it doesn’t mean no action has been taken.

Five of the moderators said it was harder to get cases reviewed or content removed if it was posted in closed Facebook groups or on Facebook Messenger. Meta “would be less strict about something that happened behind ‘closed doors,’” one team leader said. “With Messenger, we couldn’t really make a move unless the language and content was very obvious. If it was four guys who trusted each other and it was in a closed group, this situation could go on forever.” Meta said these allegations “appear to be misleading and inaccurate” and said it uses technology to find child sexualization content in private Facebook groups and on Messenger.

In 2021, Frances Haugen, a former Facebook employee who blew the whistle on the company, leaked internal documents that appear to support the moderators’ claims. These documents, thousands of pages long, detailed how the company handled allegedly criminal content. In a memo in Haugen’s leak, the company states that “Messenger groups with fewer than 32 people must be treated with full privacy protections.”

Matías Cruz*, who worked as a content moderator from 2018 to 2020, reviewing Spanish-language posts on Facebook, believes that the criteria Meta was using to recognize trafficking were too narrow to keep up with traffickers, who were constantly changing keywords to avoid detection. According to Cruz, traffickers would say, “I have this goat for sale,” and put a ridiculous price on it. Sometimes they would just say [the price] for a night or two, or ‘an hour.’” According to Cruz, it was obvious what was happening, but “managers would say the message was too vague, so they would just do nothing.”

Cruz and three other moderators said that in instances like this, where their bosses felt there was not enough evidence to escalate the case, the moderators were given lower scores for being imprecise, which in turn affected their performance evaluations. “When we were trying to help people and warn about dangers, if these situations were not investigated, we were considered to be imprecise and unable to assess the threat,” Cruz said.

Although Meta is required by law to report any images of child exploitation found on its platforms, the company is not legally responsible for crimes committed on its platform, due to a law created almost three decades ago, at the dawn of the Internet. In 1996, the US Congress passed the Communications Decency Act, which primarily aimed to ensure the regulation of pornographic content on the Internet. However, Section 230 of the act states that providers of “interactive computer services” – which include owners of social media platforms and websites – should not be treated as publishers of material posted by users. This section was included in the law to ensure the free flow of information, while protecting the growing tech industry from constant litigation.

While a newspaper, for example, has a legal obligation to prove the truth of what it publishes, Section 230 excludes liability for a company like Meta, which hosts the content of others, for what appears on its platforms. Section 230 therefore positions internet service providers as fundamentally neutral: they provide forums in which illegal, harmful, or false content can be posted and disseminated, but they are ultimately not responsible for that content. Since the law was passed, tech companies like Meta have successfully argued in US courts that Section 230 grants them immunity from prosecution for any illegal content posted on their platforms, so long as they are unaware of the existence of such content.

While a newspaper, for example, has a legal obligation to prove the veracity of what it publishes, Section 230 excludes the liability of a company like Meta, which hosts the content of others, for what appears on its platforms.​

The debate around Section 230 has become highly polarized. Those who want it changed argue that the legal safe harbor it has provided to internet companies means they have no incentive to root out illegal content from their sites. In an op-ed published in The Wall Street Journal in January, President Biden came out in favor of reforming the section. “I have long said that we must fundamentally reform Section 230,” he wrote, calling for “bipartisan action from Congress to hold Big Tech accountable.”

However, tech companies, along with internet freedom groups, argue that changes to Section 230 could lead to censorship and an erosion of privacy, particularly of private and encrypted content. These arguments about Section 230 are being tested in a landmark case that has reached the US Supreme Court and that centers on the extent to which YouTube can be held culpable for the videos it recommends to its users. A ruling is expected by the end of June.