As AI-Driven Scams Surge, Experts and Lawmakers Push for Greater Accountability From Tech Platforms
On April 30, the Institute of Global Politics (IGP) and the Center for Countering Digital Hate (CCDH) hosted lawmakers, AI experts, and researchers at Columbia University to examine how social media platforms and generative AI are fueling a surge in online scams – and what policy responses might hold tech companies accountable.
The event brought together speakers from academia, labor advocacy, and state-level policymaking who described a rapidly evolving scam ecosystem that preys on vulnerable users, especially older adults, while platforms continue to profit from fraudulent advertising
Moderated by Imran Ahmed, CEO and founder of the Center for Countering Digital Hate, the discussion connected emerging research on AI-enabled scams to New York policy proposals aimed at holding platforms accountable for the fraud they enable. Opening the event, Ahmed framed scams not as isolated mishaps but as a structural problem in the business model of major platforms such as Meta.
“Most scams only exist because platforms like Facebook can make millions from them,” said Ahmed, arguing that online scams are amplified by social media advertising and generative AI. “A scammer doesn’t even need to look for you. Platforms will find you for them, and anyone can target anyone else with extraordinary precision, which allows scammers to zero in on the perfect targets.”
He challenged the idea that online scams are simply the result of user error or bad luck. He pointed to forthcoming research from his center on Medicare Advantage scam ads running on Meta platforms, which found that scammers use the same sophisticated targeting systems as legitimate advertisers, often directing ads at older Americans.
According to CCDH, Meta has allowed scammers to inundate older Americans with misleading claims about “free benefits.” The research found that Meta permitted scam ads to generate roughly 72 million impressions before eventually removing them for violating anti-scam policies. CCDH estimated the ads generated $3.7 million in revenue before they were taken down.
Ahmed said Meta sometimes removed one version of a scam ad while allowing nearly identical versions to continue running, highlighting enforcement gaps. Citing a recent Reuters report, he also noted internal Meta estimates that roughly 10 percent of its global revenue may come from scams and illegal goods.
“Ultimately, it is clear these companies aren’t going to change until it becomes unprofitable or policymakers step in,” Ahmed said, noting New York’s proposed Fraudulent Social Media Advertising Prevention Act as one emerging response.
Alice Marwick, director of research at Data & Society Research Institute, a nonprofit organization that examines the social implications of data-driven technologies and AI, presented findings from a new report on how generative AI is changing the nature of scams worldwide. The report, conducted with Anya Schiffrin, IGP Affiliated Faculty member and codirector of SIPA’s Technology Policy and Innovation (TPI) concentration, assesses the limits of current regulatory approaches and calls for stronger institutional accountability.
“Our conclusion is basically that generative AI is making scams worse,” Marwick said. “It’s fueling a surge in scams and misinformation around the globe. Scams are more common, they’re more convincing, and they’re harder to counter.”
She cited estimates from Deloitte that Americans lost $12.3 billion to AI-related fraud in 2023, with projected losses rising to nearly $40 billion by 2027 – figures that are likely still an underestimate because many victims may never report being scammed due to shame and stigma. In Regula’s Deepfake Trends 2024 report, 92 percent of surveyed companies reported experiencing financial losses due to deepfakes.
Generative AI, Marwick explained, is increasingly being used by scammers to create convincing impersonations of celebrities, authority figures, and even family members. She pointed to fake TikTok reward programs using the likenesses of Rihanna and Taylor Swift, as well as bogus investment schemes featuring Elon Musk. “Generative AI enables highly realistic audio-visual impersonations of people from celebrities to regular folks, and this massively increases the credibility and the success rate of scams,” Marwick said. “We can’t rely on people to individually detect scams when the tools are industrialized.”
She added that the technology doesn’t just increase financial losses; it erodes public trust. “When people are scammed, it decreases their trust, not only in the people around them, but in the platforms and systems they expect to protect them,” she said.
Schiffrin placed the problem within a broader global regulatory context. Her team mapped global policy responses to deepfake financial fraud in countries including Singapore, Australia, Denmark, Ireland, Malaysia, and Mexico to identify where governments intervene in the “scam pipeline.” Approaches range from stricter advertising rules and AI disclosure requirements to limits on SIM card ownership and crackdowns on organized scam operations.
Schiffrin also highlighted the legal concept of the “cheapest cost avoider,” meaning the actor best positioned to prevent harm at the lowest cost. “One of the cheapest cost avoiders would have to be Facebook/Meta,” Schiffrin said, “because they’re the ones distributing scam ads, and they’re making money off it.”
On the policy front, New York State Senator Leroy Comrie and Assemblymember Alex Bores outlined the objectives of the Fraudulent Social Media Advertising Prevention Act, which aims to shift responsibility for detecting and preventing scam ads onto platforms.
Comrie, who recounted his own experience being scammed by a fraudulent shoe ad on Facebook, framed the bill as a basic consumer protection measure. “People shouldn’t have to navigate a digital environment where the burden is entirely dependent on them to determine what is real and what is not,” he said.
The legislation would require platforms to verify advertisers before ads go live, review ads for fraud and impersonation before publication, maintain fraud-prevention systems, retain relevant data, and provide easier mechanisms for users to report scams. It would also empower the New York Attorney General to hold platforms accountable and create a private right of action, allowing individual victims to seek redress directly.
Bores emphasized that states may need to lead in the absence of federal action. “The federal government not only doesn’t currently really have the rules to change behavior, but certainly has absolutely no willingness to do any enforcement,” he said. “Stopping scams can and should be a bipartisan issue.”
As AI-generated fakes become more convincing and platforms continue to profit from engagement regardless of authenticity, panelists stressed that public policy must shift away from individual “scam literacy” and toward systemic accountability.
That means greater transparency, stronger oversight, and legal liability for the platforms that design, operate, and monetize the systems scammers depend on. In New York, the fate of the Fraudulent Social Media Advertising Prevention Act may become an early test of how far lawmakers are willing to go in rebalancing power between users and the platforms competing for their attention.