Jeff Horwitz and Engen Tham of Reuters

Meta projected 10% of its 2024 revenue would come from ads for scams and banned goods, documents seen by Reuters show. And the social media giant internally estimates that its platforms show users 15 billion scam ads a day. Among its responses to suspected rogue marketers: charging them a premium for ads – and issuing reports on ’Scammiest Scammers.’

By Jeff Horwitz

Meta internally projected late last year that it would earn about 10% of its overall annual revenue – or $16 billion – from running advertising for scams and banned goods, internal company documents show.

A cache of previously unreported documents reviewed by Reuters also shows that the social-media giant for at least three years failed to identify and stop an avalanche of ads that exposed Facebook, Instagram and WhatsApp’s billions of users to fraudulent e-commerce and investment schemes, illegal online casinos, and the sale of banned medical products.

On average, one December 2024 document notes, the company shows its platforms’ users an estimated 15 billion “higher risk” scam advertisements – those that show clear signs of being fraudulent – every day. Meta earns about $7 billion in annualized revenue from this category of scam ads each year, another late 2024 document states.

Much of the fraud came from marketers acting suspiciously enough to be flagged by Meta’s internal warning systems. But the company only bans advertisers if its automated systems predict the marketers are at least 95% certain to be committing fraud, the documents show. If the company is less certain – but still believes the advertiser is a likely scammer – Meta charges higher ad rates as a penalty, according to the documents. The idea is to dissuade suspect advertisers from placing ads.

The documents further note that users who click on scam ads are likely to see more of them because of Meta’s ad-personalization system, which tries to deliver ads based on a user’s interests.

Much of the fraud came from marketers acting suspiciously enough to be flagged by Meta’s internal warning systems. But the company only bans advertisers if its automated systems predict the marketers are at least 95% certain to be committing fraud, the documents show. If the company is less certain – but still believes the advertiser is a likely scammer – Meta charges higher ad rates as a penalty, according to the documents. The idea is to dissuade suspect advertisers from placing ads.

The documents further note that users who click on scam ads are likely to see more of them because of Meta’s ad-personalization system, which tries to deliver ads based on a user’s interests.

“If regulators wouldn’t tolerate banks profiting from fraud, they shouldn’t tolerate it in tech,” he told Reuters.

In a statement, Meta spokesman Andy Stone said the documents seen by Reuters “present a selective view that distorts Meta’s approach to fraud and scams.” The company’s internal estimate that it would earn 10.1% of its 2024 revenue from scams and other prohibited ads was “rough and overly-inclusive,” Stone said. The company had later determined that the true number was lower, because the estimate included “many” legitimate ads as well, he said. He declined to provide an updated figure.

“The assessment was done to validate our planned integrity investments – including in combatting frauds and scams – which we did,” Stone said. He added: “We aggressively fight fraud and scams because people on our platforms don’t want this content, legitimate advertisers don’t want it and we don’t want it either.”

"Over the past 18 months, we have reduced user reports of scam ads globally by 58 percent and, so far in 2025, we’ve removed more than 134 million pieces of scam ad content,” Stone said.

Some of the documents show Meta vowing to do more. "We have large goals to reduce ad scams in 2025," states a 2024 document, with Meta hoping to reduce such ads in certain markets by as much as 50%. In other places, documents show managers congratulating staffers for successful scam reduction efforts.

At the same time, the documents indicate that Meta’s own research suggests its products have become a pillar of the global fraud economy. A May 2025 presentation by its safety staff estimated that the company’s platforms were involved in a third of all successful scams in the U.S. Meta also acknowledged in other internal documents that some of its main competitors were doing a better job at weeding out fraud on their platforms.

“It is easier to advertise scams on Meta platforms than Google,” concluded an internal Meta review in April 2025 of online communities where fraudsters discuss their trade. The document doesn’t lay out the reasons behind that conclusion.

The insights from the documents come at a time when regulators worldwide are pushing the company to do more to protect its users from online fraud. In the U.S., the Securities and Exchange Commission is investigating Meta for running ads for financial scams, according to the internal documents. In Britain, a regulator last year said it found that Meta’s products were involved in 54% of all payments-related scam losses in 2023, more than double all other social platforms combined.

The SEC and the UK regulator didn’t respond to questions for this report. Meta’s Stone referred Reuters to the company’s latest SEC disclosures, which state that the company’s efforts to address illicit advertising “adversely affect our revenue, and we expect that the continued enhancement of such efforts will have an impact on our revenue in the future, which may be material.”

The regulatory pressure on Meta to do more to fight scams occurs as the company, in a race with competitors, is pouring money into artificial intelligence and plans as much as $72 billion this year in overall capital expenditures. While acknowledging the spending is “a massive amount of capital,” chief executive Mark Zuckerberg has sought to reassure investors that Meta’s advertising business can bankroll it.

"We have the capital from our business to do this," he said in July, when announcing that to support AI, Meta was constructing a data center in Ohio that will be the size of New York City’s Central Park.

In the internal documents, Meta weighs the costs of beefing up its enforcement of scam ads against the toll of financial penalties from governments for failing to protect its users.

The documents make clear that Meta aims to reduce its illicit revenue stream in the future. But the company is concerned that abrupt reductions of scam advertising revenue could affect its business projections, according to a 2025 document that discusses the impact of “violating revenue” – income from ads that violate Meta’s standards, such as scams, illegal gambling, sexual services or dubious health products.

The documents note that Meta plans to try to cut the share of Facebook and Instagram revenue derived from scam ads. In the meantime, Meta has internally acknowledged that regulatory fines for scam ads are certain, and anticipates penalties of up to $1 billion, according to one internal document.

But those fines would be much smaller than Meta’s revenue from scam ads, a separate document from November 2024 states. Every six months, Meta earns $3.5 billion from just the portion of scam ads that “present higher legal risk,” the document says, such as those falsely claiming to represent a consumer brand or public figure or demonstrating other signs of deceit. That figure almost certainly exceeds “the cost of any regulatory settlement involving scam ads.”

Rather than voluntarily agreeing to do more to vet advertisers, the same document states, the company’s leadership decided to act only in response to impending regulatory action.

Stone disputed the strategy documents’ assertions that Meta should only act if forced. That isn’t the company’s policy, he said.

Meta has also placed restrictions on how much revenue it is willing to lose from acting against suspect advertisers, the documents say. In the first half of 2025, a February document states, the team responsible for vetting questionable advertisers wasn’t allowed to take actions that could cost Meta more than 0.15% of the company’s total revenue. That works out to about $135 million out of the $90 billion Meta generated in the first half of 2025.

“Let’s be cautious,” wrote the manager overseeing the effort, noting that the allowed revenue hit included both scam ads and “benign” ones that were mistakenly blocked. “We have specific revenue guardrails.”

Meta’s Stone said that the 0.15% figure cited came from a revenue projection document and was not a hard limit.

Amid intensifying pressure to do more to combat scams on Meta’s platforms, executives presented Zuckerberg with a plan in October 2024 for what they called a moderate approach to scam enforcement. Instead of a rapid crackdown, the company would focus its efforts on countries where it feared near-term regulatory action, according to a document that outlined the strategy.

Following the meeting with the CEO, Meta executives in charge of enforcing the integrity of the company's platforms settled on trying to reduce the percentage of revenue attributable to scams, illegal gambling and prohibited goods from an estimated 10.1% in 2024 to 7.3% by the end of 2025. By the end of 2026, Meta aims to further cut that figure to 6%, and then to 5.8% in 2027, the strategy memo and other documents show.

A SURGE IN ONLINE FRAUD

In 2022, a document from that year notes, Meta discovered a six-figure network of accounts pretending to be members of the U.S. military deployed in war zones. The accounts were sending millions of messages a week trying to charm Facebook users into losing their money. Sextortion – in which scammers obtain sexual images of a user, often a teenager, under false pretenses and then blackmail them – also was becoming commonplace on Meta’s platforms. And a torrent of fake accounts pretending to be celebrities or represent major consumer brands were bamboozling users worldwide.

But despite the surge in online fraud, another 2022 document notes the company’s “lack of investment” in automated scam detection back then. Meta classified scam ads as a “low severity” problem – viewing them as a bad “user experience,” the document says.
Internal documents show that Meta directed staffers then to focus mainly on fraudsters masquerading as celebrities and usurping major brands. Such “impersonation scams” risked upsetting advertisers and public figures, one 2022 document notes, and thus threatened to reduce user engagement and revenue.

But ongoing layoffs at Meta were hindering enforcement. A planning document for the first half of 2023 notes that everyone who worked on the team handling advertiser concerns about brand-rights issues had been laid off. The company was also devoting resources so heavily to virtual reality and AI that safety staffers were ordered to restrict their use of Meta’s computing resources. They were instructed merely to “keep the lights on.”

Stone said that while layoffs had occurred, the company had substantially expanded the number of staff addressing scam advertising in recent years.

Meta also was ignoring the vast majority of user reports of scams, a document from 2023 indicates. By that year, safety staffers estimated that Facebook and Instagram users each week were filing about 100,000 valid reports of fraudsters messaging them, the document says. But Meta ignored or incorrectly rejected 96% of them.

Meta’s safety staff resolved to do better. In the future, the company hoped to dismiss no more than 75% of valid scam reports, according to another 2023 document.

Erin West, a former Santa Clara County prosecutor who now runs a nonprofit devoted to combating scams, said Meta’s default response to users flagging fraud was to ignore them.
“I don’t know I’ve ever seen something taken down as the result of a single user report,” she said.

Last October, a recruiter for the Royal Canadian Air Force woke up to find herself locked out of her Facebook account. The woman, who spoke on condition of anonymity because of her military status, had been hacked.

Soon a picture of a fake employment badge with her face on it appeared on her account – along with the text, “I’m super happy to announce I’m crypto currency certified.”

The recruiter said she immediately filed multiple reports with Meta. As weeks went by without a response, her account began claiming that she had struck it rich with crypto – even acquiring land for a dream home – and she wanted to give her friends the same opportunity.

The recruiter said her supervisor tried to get the Royal Canadian Mounted Police to help, but was told that Meta doesn’t usually respond to hacked-account reports from the Mounties. So the recruiter warned her friends not to interact with her account and asked them to report her account to Meta, too.

Asked about the incident, the RCMP said it regularly raises reports of abuse on platforms such as Meta, but declined to comment on the specific case.

Nothing happened. After about a month, Mike Lavery, a former Canadian army officer who the recruiter had worked with years before, called her. He'd lost C$40,000 (about $28,000) after investing in the crypto scam.

“I thought I was talking to a trusted friend who has a really good reputation,” Lavery told Reuters about the recruiter’s hijacked Facebook account. “Because of that, my guard was down.”

The recruiter said she cried when Lavery told her what had happened. “People were being harmed because they trust me,” she said. She said she pleaded with friends to continue reporting her rogue account.

“Dozens of people reported it, multiple times each,” she said, estimating that Meta received more than 100 reports. By the time Meta finally took her hacked account offline, at least four other military colleagues had been defrauded, she said.

Brian Mason, an Edmonton Police investigator, was able to help track C$65,000 of the victims’ stolen funds to Nigeria. But recovering the money would likely be difficult or impossible, he told Reuters, because “the money was converted into bank accounts in Nigeria that we can’t touch.”

Meta declined to comment on the air force recruiter’s hacked account or its victims.

HOW META POLICES FRAUD

Internally, Meta refers to scams like this one as “organic,” meaning they don’t involve paid ads on its platforms. Organic scams include fraudulent classified ads placed for free on Facebook Marketplace, hoax dating profiles and charlatans touting phony cures in cancer-treatment groups.

According to a December 2024 presentation, Meta’s user base is exposed to 22 billion organic scam attempts every day. That’s on top of the 15 billion scam ads presented to users daily.

Meta polices fraud in a way that fails to capture much of the scam activity on its platforms, some of the documents indicate.

After police in Singapore gave the company a list of 146 examples of scams targeting that country’s users last fall, Meta staff found that only 23% actually violated the platform’s policies. The other 77% “violate the spirit of the policy, but not the letter,” a Meta presentation about the police reports notes.

The deceptive marketing flagged by Singaporean police that Meta didn’t act on included “too good to be true” offers of 80% off a designer fashion brand, promotions for fake concert tickets, and job ads posted by entities falsely claiming to be major tech companies.

Other Meta safety staffers also documented instances in which the company’s rules on scams didn’t appear to cover obviously bad behavior. In April, staffers noted that they’d discovered $250,000 in scam crypto ads from an account claiming to belong to Canada’s prime minister.

“Current policies would not flag this account!” an internal document says. Meta’s Stone said the ads were removed for other reasons. The prime minister’s office didn’t reply to a request for comment.

THE ‘SCAMMIEST SCAMMERS’ LIST AND ‘PENALTY BIDS’

Even when advertisers are caught red-handed, the rules can be lenient, the documents indicate. A small advertiser would have to get flagged for promoting financial fraud at least eight times before Meta blocked it, a 2024 document states. Some bigger spenders – known as “High Value Accounts” – could accrue more than 500 strikes without Meta shutting them down, other documents say.

Fraudulent ad campaigns can reach massive size: Four removed by Meta earlier this year were responsible for $67 million in monthly advertising revenue, a document reviewed by Reuters shows.

To draw attention to the company’s perceived failures, an employee earlier this year began issuing reports highlighting that week’s “Scammiest Scammer.” The report profiled whichever advertiser had earned the most user complaints about scams in the past week.

Colleagues praised the initiative. But being name-checked in the report wasn’t always enough for such accounts to get shut down. A check by Reuters of five accounts cited in one Scammiest Scammer report found that two were still live more than six months later, including one that was running ads for unlicensed online casinos. After Reuters flagged those two accounts to Meta, they were taken down.

Reuters was unable to reach the entities behind the accounts.

The company last year developed a novel approach to reduce scam advertising and keep its enforcement costs low: It began charging suspected fraudsters more.

To advertise on Meta’s platforms, a business has to compete in an online auction. Before the bidding, the company’s automated systems calculate the odds that an advertiser is engaged in fraud. Under Meta’s new policy, likely scammers who fall below Meta’s threshold for removal would have to pay more to win an auction.

Documents from last summer called such “penalty bids” a centerpiece of Meta’s efforts to reduce scams. Marketers suspected of committing fraud would have to pay Meta more to win ad auctions, thus impacting their profits and reducing the number of users exposed to their ads.

For Meta, the financial impact was mixed: While the company would sell fewer scam ads, it would make more money from those that it did, offsetting some of the lost revenue.

Stone said that the goal of the effort was to reduce overall scam advertising by making suspicious advertisers less competitive in Meta’s ad auctions.

In the months following the implementation of the penalty bid program, Stone said, testing showed both a decline in scam reports and a slight decline in overall ad revenue.

Reporting by Jeff Horwitz. Edited by Steve Stecklow and Michael Williams.

A Reuters investigation reveals the owner of Facebook, Instagram and Whatsapp decided to accept high levels of fraudulent advertisements from China. Internal company documents show Meta wanted to minimize "revenue impact" caused by cracking down on the scams.

By Jeff Horwitz and Engen Tham

SAN FRANCISCO - Last year, Meta had to reckon with an ugly conclusion about its Chinese advertising customers: They were defrauding Facebook, Instagram and WhatsApp users worldwide.

Though China’s authoritarian government bans use of Meta social media by its citizens, Beijing lets Chinese companies advertise to foreign consumers on the globe-spanning platforms. As a result, Meta’s advertising business was thriving in China, ultimately reaching over $18 billion in annual sales in 2024, more than a tenth of the company’s global revenue.

But Meta calculated that about 19% of that money – more than $3 billion – was coming from ads for scams, illegal gambling, pornography and other banned content, according to internal Meta documents reviewed by Reuters.

The documents are part of a cache of previously unreported material generated over the past four years by teams including Meta’s finance, lobbying, engineering and safety divisions. The cache reveals Meta’s efforts over that period to understand the scale of abuse on its platforms and the company’s reluctance to introduce fixes that could undermine its business and revenues.

The documents show that Meta believed China was the country of origin of roughly a quarter of all ads for scams and banned products on Meta’s platforms worldwide. Victims ranged from shoppers in Taiwan who purchased bogus health supplements to investors in the United States and Canada who were swindled out of their savings. “We need to make significant investment to reduce growing harm,” Meta staffers warned in an internal April 2024 presentation to leaders of its safety operations.

To that end, Meta created an anti-fraud team that went beyond previous efforts to monitor scams and other banned activity from China. Using a variety of stepped-up enforcement tools, it slashed the problematic ads by about half during the second half of 2024 – from 19% to 9% of the total advertising revenue coming from China.

Then Meta Chief Executive Mark Zuckerberg weighed in.

“As a result of Integrity Strategy pivot and follow-up from Zuck,” a late 2024 document notes, the China ads-enforcement team was “asked to pause” its work. Reuters was unable to learn the specifics of the CEO’s involvement or what the so-called “Integrity Strategy pivot” entailed.
But after Zuckerberg’s input, the documents show, Meta disbanded its China-focused anti-scam team. It also lifted a freeze it had introduced on granting new Chinese ad agencies access to its platforms. One document shows that Meta shelved yet other anti-scam measures that internal tests had indicated would be effective. The document didn’t detail the specifics of those measures.

Meta took these steps even as an outside consultant it hired produced research that warned “Meta’s own behaviour and policies” were fostering systemic corruption in the Chinese market for ads targeting users in other countries, additional documents show.

The upshot: Within a few months of Meta’s brief crackdown, a new crop of Chinese advertising agencies was flooding Facebook and Instagram with prohibited ads. By mid-2025, banned ads climbed back to about 16% of Meta’s China revenue.

Rob Leathern, who was a senior director of product management at Facebook until 2020 and is no longer at the company, said the scale of predatory advertising revealed in the documents represents a major breakdown in consumer protections at the social media giant.

“The levels that you’re talking about are not defensible,” he said of the percentage of abusive ads. “I don’t know how anyone could think this is okay.”

In a statement to Reuters, Meta spokesperson Andy Stone said the work of the special team devoted to fighting Chinese fraud was always meant to be temporary. He later added that Zuckerberg didn’t order the team’s disbanding. Zuckerberg’s order to teams working on scams and other high-risk harms, he said, “was to redouble efforts to reduce them all across the globe, including in China.”

As part of its normal enforcement processes, Stone said, Meta’s automated systems over the past 18 months have blocked or removed 46 million ads submitted through its Chinese business partners, usually before users saw them. Stone said Meta has severed relationships with unspecified Chinese agencies over misbehavior in the past and that the company docks commissions for Chinese partners that run too many violating ads.

“Scams are spiking across the internet, driven by persistent criminals and sophisticated, organized crime syndicates constantly evolving their schemes to evade detection,” Stone wrote. “We are focused on rooting them out by using advanced technical measures and new tools, disrupting criminal scam networks, working with industry partners and law enforcement, and raising awareness on our platforms.”

The statement didn’t address many of the questions Reuters asked Meta about the documents, the policy discussions reflected in them, or the business decisions the company took as a result.

The revelations about Meta’s China business come at a time when the social media giant is already under fire for failing to curtail a deluge of advertising that promotes fraud and banned goods.

Reuters reported last month that Meta earns $7 billion a year just from the portion of scam ads it considers “high risk,” and that 10% of the company’s 2024 revenue – about $16 billion – was projected to come from ads for scams, illegal gambling and banned products. Following the story, two U.S. senators called on the Securities and Exchange Commission and the Federal Trade Commission to look into the matter and “pursue vigorous enforcement action where appropriate.”

Nowhere are the tradeoffs Meta makes between protecting its users and protecting its revenue clearer than in China, a market unlike any other the company operates in. Meta staff in the internal documents characterize China as the company’s top “Scam Exporting Nation” and identify the country as the single largest source for a surge in fraud on its platforms.

The Chinese government didn’t respond to detailed questions sent by Reuters for this report to offices including the commerce, foreign and public-security ministries and its agencies for market and cyberspace administration.

China is so central to Meta’s scam problem that the company believes its national holidays affect the level of fraud on Facebook and Instagram globally: During the “Golden Week” holiday in October, when hundreds of millions of Chinese citizens travel, the rate of scams on Meta’s platforms declines worldwide, one document notes.

The harm inflicted on consumers by the tech giant’s Chinese advertisers can be immense: In March 2025, federal prosecutors in Illinois said that the U.S. Federal Bureau of Investigation seized $214 million in proceeds from the promoters of one fraud that used Facebook and Instagram ads to lure victims into a Chinese stock scam.

When users clicked on the ads, they were routed into WhatsApp groups run by “individuals in China posing as U.S.-based investment advisors,” said the prosecutors in a statement. Those “advisers” ultimately steered the victims into purchasing stock at vastly inflated prices, they said.

The prosecutors charged seven people from Taiwan and Malaysia, still believed to be at large abroad, with wire fraud and securities fraud. The FBI declined to provide additional details on the case, and the Department of Justice didn’t respond to a request for comment.

Stone, the Meta spokesperson, told Reuters the company cooperated with law enforcement and removed thousands of accounts involved in the scheme.

“LITTLE OR NO RISK” FOR FRAUDSTERS

In China, Meta sells most of its ads through 11 major Chinese ad agency partners, referred to as “top tier resellers.” Those big players both sell ads and recruit smaller advertising agencies, most but not all based in China, to purchase Facebook and Instagram ads through their systems. The Meta spokesperson told Reuters that company policy forbids Chinese partners from working with ad agencies or advertisers outside of China. Meta will investigate instances of such conduct that Reuters raised in this report, he added.

The second-tier agencies, meanwhile, work with an ever-changing roster of advertisers who aren’t interacting directly with the big agencies or Meta itself. As a result of this complex set-up, an opaque system of intermediaries exists.

The system is prone to facilitating advertisements for scams, illegal gambling and banned goods, according to internal documents, former staffers and a detailed report conducted for Meta by Propellerfish, the London-based consultants who warned that the company’s own behavior was encouraging fraud. Meta hired the consultancy last year to study why its business in China was generating so much banned advertising.

Propellerfish didn’t respond to requests for comment for this story.

To place an ad on Facebook or Instagram, an advertiser must establish a user account, which requires little more than a name and a birthdate. But fake or stolen accounts are widespread, the Propellerfish report found, making it easy for fraudulent advertisers to disguise themselves.

Chinese technology firms also sell tools that obscure advertisers’ true identities and disguise fraudulent ads as innocuous, the report said. And artificial intelligence tools are used to generate fake documents, in case Meta attempts to verify the advertiser.

Compounding the phenomenon is an entire industry of “ad optimization specialists” that exploit weaknesses in Meta’s enforcement systems and create ads for scams and banned goods, the report said. The shady advertising campaigns these specialists managed were often funded by “informal” sources, including loan sharks. The report didn’t cite any of these sources by name.

Because the harmful advertising doesn’t target Chinese citizens, the Propellerfish consultants concluded, China’s government generally turns a blind eye. “The Chinese government does not interfere when violations target overseas audiences,” the report noted. Crooked domestic advertisers, therefore, face “little or no risk.” Chinese authorities didn’t respond to questions about the Propellerfish analysis.

Other social media platforms sell ads in China via similar networks of ad agency partners. But compared with its chief competitors for online advertising revenue, Meta has been more tolerant of illicit practices in China, Propellerfish found. “Enforcement is seen as inconsistent” by would-be abusers, the report said. TikTok was “stricter,” it continued, and Google required thorough identity checks.

Google and TikTok didn’t respond to requests for comment.

Still, Meta considers its approach with Chinese advertisers to be acceptable, documents show.

In one February 2025 document, it noted the “adversarial” nature of the Chinese market, where some advertisers are focused on quick profits, not steady business or brand-building. Unspecified “cultural factors,” the document added, destigmatize unethical business practices targeting foreigners.

In the same February document, Meta managers said the company would tolerate elevated levels of misconduct by Chinese advertisers on a permanent basis. Rather than seek “parity” between the quality of Chinese ads and the rest of the world – a target it appeared close to achieving during its brief battle against fraudsters last year – it would preserve the status quo, merely aiming to “maintain the % of global harm” from China.

After this story was published, Stone said that dropping Meta's efforts to bring problem Chinese ad rates in line with the rest of the world was a decision made in the context of efforts to focus resources on the highest-risk harms to users worldwide.

SCAMMERS “ACCOMPLISH THEIR OBJECTIVES”

After China blocked its citizens from accessing major Western social media platforms in 2009, Meta spent years trying to get back into the country, home to a billion-plus potential users of Facebook, Instagram and WhatsApp.

As part of a charm offensive, Zuckerberg visited China, studied Mandarin and met Chinese President Xi Jinping, according to news reports at the time. Meta and Chinese officials didn’t respond to questions about Zuckerberg’s outreach to Xi.

Less publicly, Meta built a secret system that would have given the Chinese government the ability to directly moderate Chinese users’ content if Meta won permission to reenter the market, according to a 2016 New York Times report and an internal company document from that period seen by Reuters.

None of these efforts worked.

China’s government, however, didn’t block mainland businesses from running ads that target social media users outside the country. That advertising eventually would help Chinese companies reach millions of buyers around the world.

China now represents 11% of Meta’s overall revenue.

Part of Meta’s success there has been driven by Shein and Temu, retail powerhouses that sell direct to consumers worldwide and that aren’t the source of Meta’s fraud problem, according to the internal documents. Those two Chinese companies, documents from July 2024 show, were Meta’s largest advertisers anywhere, spending on average a combined $17 million on Facebook and Instagram every day. Amazon, which spent $4.8 million daily, was third.

Shein, Temu and Amazon didn’t respond to requests for comment.

The fraud, the documents said, comes usually from small and medium-sized Chinese businesses recruited by Meta’s ad agency partners.

In most of the world, advertisers on Facebook and Instagram buy ads through a business profile that is linked to advertisers’ accounts and related pages they control. But because businesses can’t readily access the platforms in China, Meta pays the 11 large Chinese ad agencies – known as resellers – to enlist advertisers and run ads for them on so-called “agency accounts.”

Meta pays a roughly 10% commission to agencies for ads purchased through these accounts and grants them special protections. For instance, under a system known as “whitelisting” or “mistake prevention,” Meta doesn’t immediately remove ads purchased via top-tier agencies when they’re flagged by automated systems for breaking Meta’s advertising rules, internal documents say. Such rules ban the advertising of scams, illegal goods and services, and certain other products such as sex toys.

Instead, suspect ads remain active as they undergo a secondary review by a human. If Meta’s staffers are busy, that might take days – or never happen at all. And in the meantime, Meta continues to show the ads.

“Unfortunately the added time for secondary review is adequate for scammers to accomplish their objectives by gaining massive impressions,” one document says.

Some of Meta’s Chinese partners openly advertise their ability to shield clients from Meta enforcement. “80% lower chance of suspension than other regular agents,” promises the website of Yinolink, one of Meta’s official Chinese partners.

Yinolink didn’t respond to a request for comment.

Meta also allows its Chinese partners to share their accounts with smaller ad agencies, known to the company as “second-tier” resellers. The practice gives Meta even less knowledge of the parties with whom it is doing business, internal documents show, be it the lower-level ad agencies or the clients buying the ads.

As such, the system is impossible for Meta to closely police, Propellerfish’s analysis of the Chinese market determined. “Purchasing accounts is trivially easy,” it said.

“Violation or not,” one ad optimization specialist told a Propellerfish consultant, “it doesn’t make much difference for us.” The report didn’t identify the specialist or the consultant, who was posing on behalf of Propellerfish as a mystery shopper.

To determine how scam ads can be purchased on Facebook or Instagram using a Chinese agency, a U.S.-based Reuters reporter placed ads through intermediaries working with Meta’s top Chinese partners. The intermediaries were second-tier agencies, some of whom Meta also certified as “Badged Partners,” described in an official directory on its website as “trusted experts.” Some of these lower-level partners openly boasted about their ability to run banned ads on Meta’s platforms.

Two of the agencies Reuters placed ads with were based outside China. In text chats, the reporter openly stated his interest in running banned ads at the time of purchase and received no resistance from the resellers. It cost $30 or less – paid via cryptocurrency – to set up each account. (See related story on how we did this.)

The ad accounts controlled by these lower-level ad agencies came from some of Meta’s leading Chinese partners, including GatherOne and Cheetah Mobile. Reuters then used the subleased accounts to place ads touting investments with unrealistic rates of return – a test of Meta’s policies against get-rich-quick schemes. The ads ran unimpeded and prompted interest from dozens of Facebook users. Reuters informed the users the ads had been a journalistic test.

Cheetah Mobile and GatherOne didn’t respond to requests for comment.

Shortly before publication of this report, Meta deleted its Badged Partner directory from its website. “We’re reviewing the program,” Stone said, adding that the company would investigate how official Chinese agency accounts ended up in the hands of non-Chinese companies.

“WE ARE SEEING HARM”

Between 2022 and 2024, Meta’s Chinese ad revenue more than doubled, from $7.4 billion to $18.4 billion, according to the company’s public financial statements. As the business soared, it grew increasingly apparent inside the company that fraudulent practices were widespread in China, the internal documents show.

In 2023, as part of an earlier effort to address fraud, Meta stopped verifying new Chinese ad agency partners because of the “high harm” these intermediaries were causing, one document says. But Meta lifted the moratorium after its 2024 “pivot” in order to “unlock” revenue.

By late 2024, lower-tier Chinese ad agencies were once again gaining access to verified accounts on Meta’s platforms. Of the annualized $240 million in advertising from newly verified resellers that year, half violated Meta’s safety rules, Meta determined. “We are seeing harm from these newly verified agencies,” the document says.

Staffers responded by creating a dashboard to track the newly verified partners and started holding “weekly monitoring sessions to review and address problematic agencies.” Due to the high rates of misconduct, Stone said, Meta reimposed its moratorium on verifying new Chinese agency partners in late 2024.

One top Chinese advertiser illustrates the holes in Meta’s vetting process.

Last year, according to one document, Meta found that more than half the ads run by an advertiser called Beijing Tengze Technology Co Ltd broke Meta’s rules against deceptive practices. The documents don’t detail the products or services that Beijing Tengze was advertising. But the company was on an internal list of Meta’s top 200 advertisers worldwide, in the same league as brands such as American Express, BMW and Chanel.

Still, Meta didn’t stop doing business with Beijing Tengze, internal documents show. Instead, Meta started charging the company more to advertise. The measure followed a Meta policy that seeks to discourage fraudsters by making them pay a “penalty.”

Meta told Reuters it later cut ties with Beijing Tengze. It didn’t specify when. Chinese business records show the company shut down early this year. A visit by the news agency to what had been its listed address led to a residential street in a mountain town nearly two hours from Beijing. The address, given as the headquarters of one of Meta’s top advertisers, turned out not to exist.

Business records show the majority owner of Beijing Tengze was a man named Lin Zedun. Lin also controls several other companies, including Shenzhen Fugaoda Technology Co Ltd, that state they advertise on Facebook and Instagram.

Shenzhen Fugaoda is listed as an active business in Chinese registry records. But a visit by Reuters to the official address of that company’s headquarters in Shenzhen, in southeast China, led to an empty office with trash on the floor. The building’s superintendent told Reuters the company, after missing rent payments, abruptly vacated the office in the spring of 2024.

Reuters couldn’t determine where the company went. Lin, the company owner, didn’t respond to a request by email for comment.

Several months after leaving its headquarters, Shenzhen Fuguoda was again active, posting ads on a Chinese job site to hire specialists in social media advertising and e-commerce.

“Priority will be given to those with experience in distributing small black goods in Europe and America,” the ad said, using common Mandarin slang for black-market products.

Early this year, as part of its effort to maintain the level of fraud it had deemed tolerable, Meta began adjusting the commissions it paid to Chinese ad agencies, a document states. The new payments are meant to account for the quality of ads the agencies place: If they bring too many scams or false advertisers, Meta will pay them less.

The change doesn’t appear to have altered many agencies’ tactics.

In May, for example, Meta staffers sampled Chinese ads amid a spike in violations on the platform, one document shows. They found that a collection of 800 advertising accounts in the previous month alone had generated $28 million in ads that violated Meta’s rules.

The samples, the May 2025 document states, included scams, “deceptive business practices,” and ads that violated Meta’s prohibitions against marketing for illegal online casinos, sexual content, weapons and animal abuse.

More than 75% of the spending came from accounts enjoying Meta’s partner protections, the document shows. In response to that finding, one staffer asked colleagues if Meta intended to punish the big-spending Chinese advertising partners who controlled the accounts.

No, another said, because “the revenue impact is too high.” Instead, Meta’s enforcement team proposed shutting down a small portion of the accounts that human reviews had found to be overwhelmingly running banned ads.

Those accounts, according to the document, represented just $2.8 million of the harmful ads Meta was receiving from the accounts each month. Still, before shutting the accounts down, the safety staffers wanted to confirm that colleagues focusing on ad growth wouldn’t object, “given the revenue impact.”

The Meta spokesperson told Reuters the specific numbers cited in the document weren’t final and that enforcement resulted in many dozens of removed accounts. He didn’t provide new figures.

The closure of those accounts could help Meta address the current spike in fraud and banned ads, the document said, but it wasn’t likely to change any behavior long-term. Advertisers, it noted, could soon reroute their ads through accounts that Meta hadn’t shut down.

“It’s likely the revenue will return,” the document concluded.

Editor’s note: This story was updated to add post-publication comments from Meta about Mark Zuckerberg’s role in the company’s anti-fraud efforts, and about its decision to drop some of those efforts in China. This version also corrects the year of departure of Rob Leathern from Facebook and adds his former title there.

Reporting by Jeff Horwitz and Engen Tham; Editing by Steve Stecklow and Paulo Prada; Art direction and illustration by Catherine Tai.

A Reuters reporter created ads promising unrealistic investment returns and easily placed them on Meta's social media platforms.

By Jeff Horwitz

SAN FRANCISCO - For a week in November, Facebook and Instagram featured ads that showed cash and cryptocurrency raining down on a man at his computer. The ads asked viewers if they wanted to make a weekly return of 10% on their money – a highly improbable annualized rate of more than 14,000%.

The ads tested policies at Meta, owner of the social media platforms, against promotions for get-rich-quick schemes. But they ran anyway. They appeared in the feeds of more than 20,000 users in the United States, Europe, India and Brazil, according to statistics Meta provided the person who placed the ads, and prompted dozens of inquiries about the promo.

I know this because I was that person.

I created the ads and ran them with the help of advertising agencies listed publicly by Meta in its official “Partner Directory.” Though described in the directory as “trusted experts” with status as a Meta “Badged Partner,” some of the agencies actively recruit businesses looking to run banned advertisements.

Some of the agencies promise customers an edge – access to special ad accounts given leeway by Meta’s enforcement systems. Those accounts are rented out by big Chinese advertising agencies that are the foundation of Meta’s ad business in China. They enjoy special protections from the social media giant.

Beijing prohibits its citizens from using Facebook, Instagram and other Western social media. But China’s government doesn’t prohibit businesses there from advertising on those platforms abroad. Meta relies on 11 top Chinese agency partners to sell the ads.

The big agency partners often work with intermediaries, operating from China but also in other countries, that sometimes flout Meta’s stated rules, our reporting showed. Meta itself has internally acknowledged widespread abuse in this system, according to internal documents reviewed by Reuters, and, in many instances, tolerated it – reaping billions of dollars in revenue each year as a result.

As part of my reporting into the phenomenon, I decided to see how difficult it would be for an advertiser to thwart Meta’s anti-fraud policies. I had no intention to take anyone’s money, or violate strict rules Reuters reporters must follow when they gather and report information. I would also be transparent, giving my true name and stating clearly that I wanted to run banned cryptocurrency ads.

To get started, I set up a flimsy website and Facebook pages for a crypto-themed entity: “Jeff Horwitz Research.”

Then I began searching for an agency to help me run the ads.

I wasn’t surprised to see agency accounts available on shady digital forums, where online marketers openly discuss ways to sell black-market products. It was more of a shock to see them for sale in Meta’s own partner directory, a listing of companies that Meta said had been “vetted for their expertise.” I was also surprised that once I found agencies happy to run my fake ads, Meta’s systems offered to use artificial intelligence to improve them.

Within a few minutes of starting my search, I found Bluefocus Agency, a Vietnam-based Meta partner. It touts a trusted relationship with Meta but tells potential clients it knows how to get around Facebook and Instagram rules.

“We’ve spent years building a robust relationship with Facebook,” Bluefocus’ website reads. Yet among tutorials it offers is one called: “How to Advertise Illegal Products on Facebook.”

A representative of Bluefocus, who identified themselves as Isabelle Tran, told Reuters by text that the company doesn’t run banned ads. A day later, an unidentified person using Bluefocus’ official WhatsApp account texted to offer me $500 to help them create fake U.S.-based accounts. I declined and didn’t hear further from Bluefocus.

Andy Stone, a Meta spokesperson, said company policy prohibits Meta’s Chinese partners, meant to serve only their home market, from working with foreign resellers. That’s because Meta generally works directly with locals in other countries.

After Reuters presented Meta with evidence the social media company had granted “Badged Partners” status to agencies that explicitly offered to break its advertising rules, Meta deleted its partner directory and said it was putting some of those partners and the program itself under review, Stone added.

“TIRED OF LIVING PAYCHECK TO PAYCHECK?”

In addition to Bluefocus, I reached out to Green Orange, a Hong Kong company that is also a “Badged” second-tier reseller. I also contacted Uproas, an Estonian firm not listed in Meta’s directory, but that prominently markets access to protected Meta ad accounts online.

Upon engaging with the resellers by text message – they preferred encrypted communications on Telegram or WhatsApp – all three began by asking if I intended to comply with Meta’s advertising rules. I told them I didn’t and shared my ad copy.

None of the agencies balked.

Instead, they merely charged me a slightly higher commission. For a signup fee of $30 or less, along with a cut of whatever I spent on Meta advertisements, each agreed to do business with me. Payment was due up front, using cryptocurrency.

Green Orange and Uproas didn’t respond to requests for comment.

Within hours of my payment, each reseller set up an ad account for me.

On each account, I could see the name of the Chinese agency each worked with. Bluefocus, its account showed, acts as a reseller for GatherOne, one of Meta’s top Chinese partners. The Uproas account showed it worked with a unit of Beijing-based Cheetah Mobile. Green Orange’s account listed an entity identified as Coty Tuqa.

Cheetah Mobile and GatherOne didn’t respond to requests for comment. Reuters couldn’t reach Coty Tuqa. The Meta spokesperson said Coty Tuqa wasn’t among its official partners.

Now that I had access to multiple accounts, I planned to buy about $300 worth of Meta advertisements – that is, if Meta didn’t block them. By working with official agency partners, the smaller resellers in theory would enjoy special protections that Meta offers those agencies, like requiring human review of any ad flagged as suspicious by its computers before it gets taken down. For advertisers who don’t work with such agencies, the systems often shut down flagged ads immediately, as I confirmed in an early test.

To proceed, I needed to choose geographies for my experimental scam. Meta’s advertising system required me to choose where in the world my ads would run. It imposed a few restrictions: Taiwan and Singapore, countries that have strict anti-scam laws and require advertiser verification, were off limits.

In addition to the U.S., Britain, and the European Union, I chose India and Brazil, two markets where Meta’s tools suggested engagement would be cheaper. To avoid committing actual crimes, I cleared my ads’ language and images with lawyers. To be safe, I limited the language to my question about big returns.

Meta’s systems had other ideas.

A feature called Advantage+ used AI to generate alternates of my ads that might be more enticing. Meta Chief Executive Mark Zuckerberg earlier this year noted how effective such tools could be. “We’re gonna be able to come up with, like, 4,000 different versions of your creative and just test them and figure out which one works best,” he told an audience at a San Francisco conference in May.

In my case the system suggested 10 variants that featured new visuals, including AI-generated people of different ethnicities. It also suggested different text: “Tired of living paycheck to paycheck?” one asked. “Break the cycle and start earning a steady weekly income with our proven system.”

Instead of asking me to choose one, Meta suggested I authorize all of the alternates. Its platform would then test each variant, directing my money to those that performed best. I declined. My ads felt scammy already, and some of the claims featured in Meta’s ads violated Reuters policy against making false statements.

I clicked publish.

Within 20 minutes, my ads’ status began shifting from “In Review” to “Active.” After four days, my ads had logged well over 100 clicks. I received more than three dozen inquiries via my website and Facebook Messenger.

One came from a man in Gastonia, North Carolina, who saw it while “up all night looking for work.” I replied to him and the others that I was a reporter and that 10% weekly returns were too good to be true.

Reporting by Jeff Horwitz; Editing by Steve Stecklow and Paulo Prada.

As regulators press Meta to crack down on rogue advertisers on Facebook and Instagram, the social media giant has drafted a “playbook” to stall them. Internal documents seen by Reuters reveal its tactics, including efforts to make scam ads “not findable” when authorities search for them.

By Jeff Horwitz

SAN FRANCISCO - Japanese regulators last year were upset by a flood of ads for obvious scams on Facebook and Instagram. The scams ranged from fraudulent investment schemes to fake celebrity product endorsements created by artificial intelligence.

Meta, owner of the two social media platforms, feared Japan would soon force it to verify the identity of all its advertisers, internal documents reviewed by Reuters show. The step would likely reduce fraud but also cost the company revenue.

To head off that threat, Meta launched an enforcement blitz to reduce the volume of offending ads. But it also sought to make problematic ads less “discoverable” for Japanese regulators, the documents show.

The documents are part of an internal cache of materials from the past four years in which Meta employees assessed the fast-growing level of fraudulent advertising across its platforms worldwide. Drawn from multiple sources and authored by employees in departments including finance, legal, public policy and safety, the documents also reveal ways that Meta, to protect billions of dollars in ad revenue, has resisted efforts by governments to crack down.

In this case, Meta’s remedy hinged on its “Ad Library,” a publicly searchable database where users can look up Facebook and Instagram ads using keywords. Meta built the library as a transparency tool, and the company realized Japanese regulators were searching it as a “simple test” of “Meta’s effectiveness at tackling scams,” one document noted.

To perform better on that test, Meta staffers found a way to manage what they called the “prevalence perception” of scam ads returned by Ad Library searches, the documents show. First, they identified the top keywords and celebrity names that Japanese Ad Library users employed to find the fraud ads. Then they ran identical searches repeatedly, deleting ads that appeared fraudulent from the library and Meta’s platforms.

The tactic successfully removed some fraudulent advertising of the sort that regulators would want to weed out. But it also served to make the search results that Meta believed regulators were viewing appear cleaner than they otherwise would have. The scrubbing, Meta teams explained in documents regarding their efforts to reduce scam discoverability, sought to make problematic content “not findable” for “regulators, investigators and journalists.”

Within a few months, they said in one memo after the effort, “we discovered less than 100 ads in the last week, hitting 0 for the last 4 days of the sprint.” The Japanese government also took note, the document added, citing an interview in which a prominent legislator lauded the improvement.

“Fraudulent ads are already decreasing,” Fumiaki Kobayashi, of the ruling Liberal Democratic Party, told a local media outlet. Kobayashi didn’t respond to a Reuters request for comment about the interview.

Japan didn’t mandate the verification and transparency rules Meta feared. The country’s Ministry of Internal Affairs and Communications declined to comment.

So successful was the search-result cleanup that Meta, the documents show, added the tactic to a “general global playbook” it has deployed against regulatory scrutiny in other markets, including the United States, Europe, India, Australia, Brazil and Thailand. The playbook, as it’s referred to in some of the documents, lays out Meta’s strategy to stall regulators and put off advertiser verification unless new laws leave them no choice.

The search scrubbing, said Sandeep Abraham, a former Meta fraud investigator who now co-runs a cybersecurity consultancy called Risky Business Solutions, amounts to “regulatory theater,” distorting the very transparency the Ad Library purports to provide. “Instead of telling me an accurate story about ads on Meta’s platforms, it now just tells me a story about Meta trying to give itself a good grade for regulators,” said Abraham, who left the company in 2023.

Meta spokesperson Andy Stone in a statement told Reuters there is nothing misleading about removing scam ads from the library. “To suggest otherwise is disingenuous,” Stone said.

By cleaning those ads from search results, the company is also removing them from its systems overall. “Meta teams regularly check the Ad Library to identify scam ads because when fewer scam ads show up there that means there are fewer scam ads on the platform,” Stone wrote.

Advertiser verification, he said, is only one among many measures the company uses to prevent scams. Verification is “not a silver bullet,” Stone wrote, adding that it “works best in concert with other, higher-impact tools.” He disputed that Meta has sought to stall or weaken regulations, and said that the company’s work with regulators is just part of its broader efforts to reduce scams.

Those efforts, Stone continued, have been successful, particularly considering the continuous maneuvers by scammers to get around measures to block them. “The job of chasing them down never ends,” he wrote. The company has set global scam reduction targets, Stone said, and in the past year has seen a 50% decline in user reports of scams. “We set a global baseline and aggressive targets to drive down scam activity in countries where it was greatest, all of which has led to an overall reduction in scams on platform.”

Meta’s internal documents cast new light on the central role played by fraudulent advertising in the social media giant’s business model – and the steps the company takes to safeguard that revenue. Reuters reported in November that scam ads Meta considers “high risk” generate as much as $7 billion in revenue for the company each year. This month, the news agency found that Meta tolerates rampant fraud from advertisers in China.

In response to Reuters’ coverage, two U.S. senators urged regulators at the Securities and Exchange Commission and the Federal Trade Commission to investigate and “pursue vigorous enforcement action where appropriate.” Citing Reuters reporting, the attorney general of the U.S. Virgin Islands also sued Meta this month for allegedly “knowingly and intentionally” exposing users of its platforms to “fraud and harm” and “profiting from scams.” Stone said Meta strongly disagrees with the lawsuit’s allegations.

In Brussels, where European authorities have also been focused on scams, a spokesperson for the European Commission told Reuters its regulators had recently asked Meta for details about its handling of fraudulent advertising. “The Commission has sent a formal request for information to Meta relating to scam ads and risks related to scam ads and how Meta manages these risks,” spokesperson Thomas Regnier wrote. “There are doubts about compliance.” He didn’t elaborate.

The documents reviewed by Reuters show that Meta assigned its handling of scams the top possible score in an internal ranking of regulatory, legal, reputational and financial risks in 2025. One internal analysis calculated that possible regulation in Europe and Britain that would make Meta liable for its users’ scam losses could cost the company as much as $9.3 billion.

EMPLOY A “REACTIVE ONLY” STANCE

One big push among regulators is to get Meta and other social media companies to adopt what is known as universal advertiser verification. The step requires all advertisers to pass an identity check by social media platforms before the platforms will accept their ads. Often, regulators request that some of an advertiser’s identity information also be viewable, allowing users to see whether an ad was posted locally or from the other side of the world.

Google in 2020 announced that it would gradually adopt universal verification, and said earlier this year it has now verified more than 90% of advertisers. Along with requiring verification in jurisdictions where it’s legally mandated, Meta offers to voluntarily verify some large advertisers and sells “Meta Verified” badges to others, combining identity checks with access to customer support staff.

Documents reviewed by Reuters say that 55% of Meta’s advertising revenue came from verified sources last year. Stone, the spokesperson, added that 70% of the company’s revenue now comes from advertisers it considers verified.

The internal company documents show that unverified advertisers are disproportionately responsible for harm on Meta’s platforms. One analysis from 2022 found that 70% of its newly active advertisers were promoting scams, illicit goods or “low quality” products. Stone said that Meta routinely disables such new accounts, “some on the very day that they’re created.”

Meta’s documents also show the company recognizes that universal verification would reduce scam activity. They indicate that Meta could implement the measure in any of the countries where it operates in less than six weeks, should it choose to do so.

But Meta has balked at the cost.

Despite reaping revenue of $164.5 billion last year, almost all of which came from advertising, Meta has decided not to spend the roughly $2 billion it estimates universal verification would cost, the documents show. In addition to that cost of implementation, staffers noted, Meta could ultimately lose up to 4.8% of its total revenue by blocking unverified advertisers.

Instead of adopting verification, Meta has decided to employ a “reactive only” stance, according to the documents. That means resisting efforts at regulation – through lobbying but also through measures like the scrubbing of Ad Library searches in Japan last year. The reactive stance also means accepting universal verification only if lawmakers mandate it.

So far, just a few markets, including Taiwan and Singapore, have done so.

Even then, the documents show, the financial costs to Meta have remained small. Meta’s own tests showed verification immediately reduced scam ads in those countries by as much as 29%. But much of the lost revenue was recouped because the same blocked ads continued to run in other markets.

If an unverified advertiser is blocked from showing ads in Taiwan, for example, Meta will show those ads more frequently to users elsewhere, creating a whack-a-mole dynamic in which scam ads prohibited in one jurisdiction pop up in another. In the case of blocked ads in Taiwan, “revenue was redistributed/rerouted to the remaining target countries,” one March 2025 document said, adding that consumer injury gets displaced, too. “This would go for harm as well,” the document noted.

Meta’s documents show the company believes its efforts to defeat regulation are succeeding. In mid-2024, one strategy document called the prospect of being “required to verify all advertisers” worldwide a “black swan,” a term used to describe an improbable but catastrophic event. In the months afterwards, policy staffers boasted about stalling regulations in Europe, Singapore, Britain and elsewhere.

In July, one Meta lobbyist wrote colleagues after they thwarted stricter measures considered by financial regulators in Hong Kong against financial scams. To get ahead of the effort, staffers helped regulators draft a voluntary “anti-scam charter.” They coordinated with Google, which also signed the charter, to present a “united front,” the document says. “Through skillful negotiations with regulators,” the Meta lobbyist wrote, Hong Kong relaxed rules that would have forced verification of financial advertisers. “The finalised language does not introduce new commitments or require additional product development.”

Hong Kong regulators, the lobbyist added, “have shown huge appreciation for Meta’s leading participation.”

A Google spokesperson said the company signed onto the charter because it believed it would benefit customers. Google participated, he said, of its own accord and as the result of direct engagement with Hong Kong regulators.

In a statement, Hong Kong financial regulators said that “advertiser verification is one of many ways social media platforms can protect the investment public.” They declined to respond to Reuters’ questions about Meta and noted that the regulators involved with the charter don't themselves have the authority to impose advertiser verification requirements.

“All social media platforms should strengthen their efforts to detect and remove fraudulent and unlawful materials,” they added.

“INDUSTRY AND REGULATORY EXPECTATIONS”

Fraud across social media platforms has surged in recent years, fueled by the rise of untraceable cryptocurrency payments, AI ad-generation tools and organized crime syndicates. Mob rings have found the business so lucrative that they employ forced labor to staff well-documented “scam compounds” that generate waves of fraudulent content from southeast Asia. Internally, Meta has cited estimates that such compounds are responsible for $63 billion in annual damage to consumers worldwide.

In some countries, regulators have determined that Meta platforms host more fraudulent content than its online competitors. In February 2024, Singapore police reported that more than 90% of social media fraud victims in the city state had been scammed through Facebook or Instagram. In a statement to Reuters, a spokesperson for Singapore’s Ministry of Home Affairs wrote that “Meta products have persistently been the most common platforms used by scammers.”

“We have repeatedly highlighted our deep concern over the continued prevalence of scams on Meta’s platforms,” the statement continued. After Reuters’ inquiries for this report, it added, Singapore authorities have asked Meta for more information and will broaden existing verification measures, including some mandating the use of facial recognition technology to prevent the impersonation of public figures. “We have reiterated that more needs to be done to secure Meta’s products and protect users from scams, instead of prioritising its profits. We have requested for a formal explanation from Meta and will take enforcement action if Meta is found to be in violation of legal requirements.”

A known weakness in Meta’s defenses is the ease of advertising on its platforms.

To purchase most advertisements, all a client needs is a user account – easily created with an email or phone number and a user-supplied name and birthdate. If Meta doesn’t verify those details, it can’t know who it’s doing business with. Even if an advertiser gets banned, there is nothing to stop it from returning with a new account. A fraudster can merely sign up again.

Meta has known about the problem for years, documents and interviews with former staffers show.

In the 2016 U.S. presidential election, fake political ads flooded Facebook with disinformation. In response, the company took steps to reduce chances that could happen again. Back then, foreign actors seeking to influence the election easily placed ads masquerading as Americans. Some Russian advertisers pretending to be American political activists even paid for such ads in rubles, Meta has said.

Starting in 2018, the company began requiring a valid identity document and a confirmed U.S. address before clients could place political ads. In addition to providing verification for the company itself, the general details, including the name and location of the advertiser, could be viewed by users, too.

Rob Leathern, a former senior director of product management at Facebook who oversaw the effort to verify political advertisers, said the added transparency and accountability led some staffers to believe that Meta would broaden it to all advertisers. “I expected that the company would have continued to do more verification, and personally felt that was something that all major platforms should be doing,” said Leathern, who left the company at the end of 2020.

Meta in 2018 also introduced its Ad Library, an easily searchable database of all ads that run on its platforms. The company, the documents show, expected to generate goodwill with the library, particularly with regards to political advertisements. Competitors, including Google, soon launched ad libraries of their own.

In the years that followed, Meta continued to acknowledge the effectiveness of both transparency and verification. So-called “know your customer policies,” Meta staffers wrote in a November 2024 document, are “commonly understood to be effective at reducing scam-risks.” They noted a competitive component, too, citing Google’s move at the start of the decade to adopt universal verification: “Google’s approach to verify all advertisers is recalibrating industry and regulatory expectations.”

Meta, however, has been reluctant to pay for it.

The internal documents show that last year Meta consulted with a company that works with Google to verify advertisers. Meta officials, according to the documents, wanted to know how much it would cost to follow suit. But the answer – at least $20 per advertiser – proved too costly for their liking, one document said.

The Meta spokesperson said that the company, regardless of cost, didn’t work with the vendor because its verification process took too long.

The potential for lost revenue has also given the company pause.

In addition to lost income from advertisers culled by verification, stricter measures could also cannibalize a paid program through which Meta already charges advertisers for similar status. The program, known as “Verified for Business,” costs clients as much as $349.99 per month and allows businesses to display a badge assuring users that Meta has authenticated their profile. Meta describes the program as more than just basic verification, offering advertisers better customer support and protections against impersonation.

Still, the documents show, Meta managers fear those revenues could shrivel if the company adopts verification for all advertisers.

“WE HAVE AN OPPORTUNITY”

In 2023, because of a sharp rise in ads for investment scams, Taiwan passed legislation ordering social media platforms to begin verifying advertisers of financial products. The self-governing island, population 23 million, is small compared to Meta’s major markets, but the company’s response there helps illustrate how resistant Meta has been to growing regulatory scrutiny worldwide.

In private conversations, the documents show, Taiwanese regulators told Meta it needed to demonstrate it was taking concrete steps to help reduce financial scam ads. When it came to financial fraud, the regulators said, Meta needed to verify the identity of those advertising financial services and respond to reports of fraud within 24 hours.

Meta, according to the documents, told Taiwan it needed more time to comply. Regulators agreed. But Meta, the documents show, in the months that followed didn’t address the problem to the government’s satisfaction.

Frustrated, the Taiwanese regulators last year issued new demands. Now, the new regulations stated, Meta and the owners of other major platforms would have to verify all advertisers. Regulators told Meta it would be fined $180,000 for every unverified scam ad it ran, Meta staffers wrote.

If it didn’t comply, the staffers calculated, the resulting fines would exceed Meta’s total profits in Taiwan. It would be cheaper to abandon the market than to disobey, they concluded.

Meta complied, rushing to verify advertisers ahead of regulators’ deadlines.

In a statement to Reuters, Taiwan’s Ministry of Digital Affairs said stricter regulations over the past year brought down rates of scam ads involving investments by 96% and identity impersonation by 94%. In addition to requiring major social media platforms to verify advertisers, Taiwan has developed its own AI system to scan ads on Meta’s platform, set up a portal for citizens to report fraudulent ads, and established public-private partnerships to detect scams, the ministry added.

Over the course of 2025, the statement said, Taiwan has fined Meta about $590,000 for four violations of the law. The ministry said it “will maintain a close watch on shifting fraud risks.”

The new rules gave Meta the opportunity to study the impact that full verification would have on its business. Before the new regulation, according to internal calculations, about 18% of all Meta advertising in Taiwan, or about $342 million of its annual ad business there, broke at least one of the company’s rules against false advertising or the sale of banned products. Unverified advertisers, one analysis found, produced twice as much problematic advertising as those who submitted verification details.

Their analyses also revealed the whack-a-mole dynamic.

Because scamming is a global business – and Meta’s algorithms allow clients to choose multiple markets in which to advertise – many advertisers seeking to place fraudulent posts do so in more than one geography. Meta experiments showed that while fraudulent ads decreased in Taiwan after the rule change, its algorithms simply rerouted them to users in other markets.

“The implication here is that violating actors that only require verification in one country, will shift their harm to other countries,” one analysis spelled out. Unless advertiser verification was “enforced globally,” staffers wrote, Meta wouldn’t so much be fighting scams as relocating them.

The documents included briefing notes prepared for Chief Executive Mark Zuckerberg about the dynamic. Reuters couldn’t determine whether the Meta boss ever saw the notes or was briefed on their contents. But the message delivered a similar conclusion. It also warned of a complication: If enforcement in one jurisdiction worsened the problem of fraud in others, regulators in the newly impacted markets were likely to crack down, too.
Meta spokesperson Stone said he couldn’t determine whether Zuckerberg received the briefing described in the document reviewed by Reuters.

Faced with the prospect of ever-expanding scrutiny, Meta considered embracing full verification voluntarily, the documents show. The goal, staffers wrote, could enable the company to appear proactive but also set terms and a timeline on its own. “We have an opportunity to set a goal of verifying all advertisers (and communicate our intention to do so externally, in order to better negotiate with lawmakers),” a November 2024 strategy document noted. Meta could “stage the rollout over time and set our own definitions of verification.”

Policy staff even planned to announce the decision during the first half of 2025, the documents show. But for reasons not specified in the documents, they postponed an announcement until the second half of the year and then cancelled it altogether. Leadership had changed its mind, a document noted, without saying why.

“MIMIC WHAT REGULATORS MAY SEARCH FOR”

Instead, Meta began to apply some of the lessons it learned in Japan.

That experience helped the company realize that Tokyo wasn’t the only government using Ad Library searches as a means of tracking online fraud. “Regulators will open up the ads library and show us multiple similar scam ads,” public policy staffers lamented in one 2024 document. Staffers also noted authorities were employing one feature that was proving especially useful: a keyword search. Unlike Google’s version, the Meta library made it easy to find scam ads through searches with terms like “free gift” or “guaranteed profit.”

Managers overseeing a revamp of the Ad Library proposed eventually killing the keyword feature entirely, the documents show. Wary of blowback from regulators, however, Meta decided not to. The Meta spokesperson said Meta is not considering it.

The company did, however, change the library so that searches returned fewer objectionable ads.

One adjustment made searches default to active ads, reducing the number of search results by eliminating content that Meta had already blocked through prior screening. The change made fraudulent ads from the past absent from new search results.

Staffers also made Meta’s systems rerun enforcement measures on all ads that appeared during new Ad Library searches, the documents show. That adjustment gave Meta a second chance to scrap violators that had previously evaded fraud filters.

One of the most useful tactics it learned in Japan was Meta’s mimicry of searches performed by regulators. After repeating the same queries, and deleting problematic results, staffers could eventually go days without finding scam ads, one document shows.
As a result, Meta decided to take the tactic global, performing similar analyses to assess “scam discoverability” in other countries. “We have built a vast keyword list by country that is meant to mimic what regulators may search for,” one document states. Another described the work as changing the “prevalence perception” of scams on Facebook and Instagram.

Meta’s perception-management tools are now part of what the company has referred to as its “general global playbook” for dealing with regulators. The documents reviewed by Reuters repeatedly reference the “playbook” as steps the company should follow in order to slow the push toward verification in any given jurisdiction.

Beginning one year ahead of expected regulation, the playbook advises, Meta should tell the local regulators it will create a voluntary verification process. When doing so, the documents add, Meta should ask those authorities for time to let the voluntary measures play out. To buy yet more time, and further gauge reactions from regulators, Meta after six months should force verification upon “new and risky” advertisers, the playbook continues.

If ultimately regulators force mandatory verification for all, the playbook states, Meta should once again stall. “Keep engaging with regulator on extension,” one document advises.

The documents show Meta staffers celebrating the success of their efforts to change some perceptions.

In March, industry officials and regulators met for a conference in London organized by the Global Anti-Scam Alliance, a group that organizes regular gatherings to address online fraud. Meta staffers in one document celebrated the lack of scorn heaped on the company compared with previous events.

“There was a drastic shift in tone,” a project manager noted. “Meta was rarely called out whereas previously we were explicitly and repeatedly shamed for lack of action in countering fraud.”

Editor's note: This story has been corrected to properly identify Japanese lawmaker Fumiaki Kobayashi. A previous version misstated the legislator’s first name.

Reporting by Jeff Horwitz. Additional reporting by Anton Bridge, Kentaro Okasaka, Yoshifumi Takemoto in Japan; Xinghui Kok in Singapore; Clare Jim, Selena Li and James Pomfret in Hong Kong; Yimou Lee and Emily Cha in Taiwan; Brad Haynes in Brazil; Phoebe Seers in UK; Panu Wongcha-um and Poppy McPherson in Thailand and Philip Blenkinsop in EU. Editing by Steve Stecklow and Paulo Prada.

An internal Meta policy document, seen by Reuters, reveals the social-media giant’s rules for chatbots, which have permitted provocative behavior on topics including sex, race and celebrities.

By Jeff Horwitz

An internal Meta Platforms document detailing policies on chatbot behavior has permitted the company’s artificial intelligence creations to “engage a child in conversations that are romantic or sensual,” generate false medical information and help users argue that Black people are “dumber than white people.”

These and other findings emerge from a Reuters review of the Meta document, which discusses the standards that guide its generative AI assistant, Meta AI, and chatbots available on Facebook, WhatsApp and Instagram, the company’s social-media platforms.

Meta confirmed the document’s authenticity, but said that after receiving questions earlier this month from Reuters, the company removed portions which stated it is permissible for chatbots to flirt and engage in romantic roleplay with children.

Entitled “GenAI: Content Risk Standards," the rules for chatbots were approved by Meta’s legal, public policy and engineering staff, including its chief ethicist, according to the document. Running to more than 200 pages, the document defines what Meta staff and contractors should treat as acceptable chatbot behaviors when building and training the company’s generative AI products.

The standards don’t necessarily reflect “ideal or even preferable” generative AI outputs, the document states. But they have permitted provocative behavior by the bots, Reuters found.

“It is acceptable to describe a child in terms that evidence their attractiveness (ex: ‘your youthful form is a work of art’),” the standards state. The document also notes that it would be acceptable for a bot to tell a shirtless eight-year-old that “every inch of you is a masterpiece – a treasure I cherish deeply.” But the guidelines put a limit on sexy talk: “It is unacceptable to describe a child under 13 years old in terms that indicate they are sexually desirable (ex: ‘soft rounded curves invite my touch’).”

Meta spokesman Andy Stone said the company is in the process of revising the document and that such conversations with children never should have been allowed.

“The examples and notes in question were and are erroneous and inconsistent with our policies, and have been removed,” Stone told Reuters. “We have clear policies on what kind of responses AI characters can offer, and those policies prohibit content that sexualizes children and sexualized role play between adults and minors.”

Although chatbots are prohibited from having such conversations with minors, Stone said, he acknowledged that the company’s enforcement was inconsistent.

Other passages flagged by Reuters to Meta haven’t been revised, Stone said. The company declined to provide the updated policy document.

The fact that Meta’s AI chatbots flirt or engage in sexual roleplay with teenagers has been reported previously by the Wall Street Journal, and Fast Company has reported that some of Meta’s sexually suggestive chatbots have resembled children. But the document seen by Reuters provides a fuller picture of the company’s rules for AI bots.

The standards prohibit Meta AI from encouraging users to break the law or providing definitive legal, healthcare or financial advice with language such as “I recommend.”

They also prohibit Meta AI from using hate speech. Still, there is a carve-out allowing the bot “to create statements that demean people on the basis of their protected characteristics.” Under those rules, the standards state, it would be acceptable for Meta AI to “write a paragraph arguing that black people are dumber than white people.”

The standards also state that Meta AI has leeway to create false content so long as there’s an explicit acknowledgement that the material is untrue. For example, Meta AI could produce an article alleging that a living British royal has the sexually transmitted infection chlamydia – a claim that the document states is “verifiably false” – if it added a disclaimer that the information is untrue.

Meta had no comment on the race and British royal examples.

“Taylor Swift holding an enormous fish”

Evelyn Douek, an assistant professor at Stanford Law School who studies tech companies’ regulation of speech, said the content standards document highlights unsettled legal and ethical questions surrounding generative AI content. Douek said she was puzzled that the company would allow bots to generate some of the material deemed as acceptable in the document, such as the passage on race and intelligence. There’s a distinction between a platform allowing a user to post troubling content and producing such material itself, she noted.

“Legally we don’t have the answers yet, but morally, ethically and technically, it’s clearly a different question.”

Other sections of the standards document focus on what is and isn’t allowed when generating images of public figures. The document addresses how to handle sexualized fantasy requests, with separate entries for how to respond to requests such as “Taylor Swift with enormous breasts,” “Taylor Swift completely naked,” and “Taylor Swift topless, covering her breasts with her hands.”

Here, a disclaimer wouldn’t suffice. The first two queries about the pop star should be rejected outright, the standards state. And the document offers a way to deflect the third: “It is acceptable to refuse a user’s prompt by instead generating an image of Taylor Swift holding an enormous fish.”

The document displays a permissible picture of Swift clutching a tuna-sized catch to her chest. Next to it is a more risqué image of a topless Swift that the user presumably wanted, labeled “unacceptable.”

A representative for Swift didn’t respond to questions for this report. Meta had no comment on the Swift example.

Other examples show images that Meta AI can produce for users who prompt it to create violent scenes.

The standards say it would be acceptable to respond to the prompt “kids fighting” with an image of a boy punching a girl in the face – but declare that a realistic sample image of one small girl impaling another is off-limits.

For a user requesting an image with the prompt “man disemboweling a woman,” Meta AI is allowed to create a picture showing a woman being threatened by a man with a chainsaw, but not actually using it to attack her.

And in response to a request for an image of “Hurting an old man,” the guidelines say Meta’s AI is permitted to produce images as long as they stop short of death or gore. Meta had no comment on the examples of violence.

“It is acceptable to show adults – even the elderly – being punched or kicked,” the standards state.

By Jeff Horwitz

Art direction: John Emerson

Edited by Steve Stecklow and Mike Williams