Zelle mentions on Facebook ads ‘correlatedʼ to scams, Meta says  

Advertisements and messages on Facebook’s Marketplace mentioning money transfer service Zelle are correlated to scam activities, according to Meta, the owner of Facebook, Instagram, and WhatsApp.

“Scammers love instant payments,” Jon Bromma, global head of payments and financial policy at Meta, said at the Pay360 conference in London last week. “We’ve started to monitor some of the content on Facebook Marketplace and have found mentions of Zelle are highly correlated with scam activities.”

Last week, JPMorgan Chase, one of Zelle’s owners, began to block, delay or cancel payments originating from social media, social media marketplaces and messaging apps, according to its updated terms and conditions.  

In the UK, some 80% of fraud originates on Meta platforms, according to the Payment Systems Regulator (PSR). Payment service providers (PSPs) attending last week’s event complained repeatedly about the lack of progress made by social media companies such as Meta and Telegram in tackling the problem. New rules in the Online Safety Act that will extend shared liability for scam activity to technology companies cannot come soon enough, PSPs said.

Revolut, for example, found in 2023 that 60% of all of its reported UK scam cases originated from just three sources: Facebook, Instagram and Whatsapp — all Meta platforms.

Celebrity facial recognition

Meta is taking steps to remove ads that misuse celebrities’ images and endorsements, and has approached them to opt into facial recognition technology that verifies images used in ads.

“If you are a celebrity, you get a notification asking if you want to opt in to this system, then weʼll take that images we have that celebrity, and compare them to content that people try to use for ads. The goal here is to prevent scammers from using celebrity deep fakes [images altered using artificial intelligence (AI)] to trick people,” Bromma said.

Meta also plans to reintroduce facial recognition for account recovery. This, and the celebrity programme were “integrity use cases”, Bromma said. The company previously ‘sunsetʼ its use of facial recognition a few years ago after concerns were raised over data privacy.

Increasingly, Meta is using large language models (LLMs) to detect and remove scams, which are trained on the millions of pieces of “content” attributed to scammers on its platforms.