How AI Accelerates Disaster Fraud - The Margin

During the middle of summer in Puerto Rico this year, amid the sweltering tropical heat, residents of the island found themselves navigating a severe water crisis – the result of a historic drought, exacerbated by deteriorating infrastructure and persistent government mismanagement.

As half a million residents were rationing their water, government agencies and humanitarian groups mobilized to provide relief by distributing water on the ground.

By early August, good Samaritans seeking to donate money to support these efforts could find the website ProjectPRRelief.org circulating on the social media app Threads, where users posted screenshots of their donations to the charity. 

That’s when Venessa Gonzalez saw it. The resident of Corozal, Puerto Rico, was skeptical: “I find it strange that not a single one of my people that is ACTIVE in community in Puerto Rico is connected in ANY way or knows someone to projectprrelief.org,” she posted on August 11.

According to an archive of the Project PR Relief website, the organization was founded in 2017, after Hurricane Maria, by a Puerto Rican living in the U.S. A timeline on the organization’s website said it  was “formally registered” in 2020 and that it serves all 78 of Puerto Rico’s municipalities. 

At first glance, the website looks legitimate. But closer investigation from The Margin revealed a flurry of red flags. The site was designed using generative artificial intelligence. The Employer Identification Number previously listed on its website shows no results on the IRS database of tax-exempt organizations, which includes charities and nonprofits. Nor can it be found in Puerto Rico’s Department of State entity database. There are no actual people — no named employees, volunteers, or founders — listed anywhere on the site.

With such limited information about the relatively new organization, it’s hard to say definitively whether the group is a scam, according to Bhupendra Acharya, an assistant professor in the University of Louisiana at Lafayette School of Computing and Informatics who studies cybercrime and AI-driven threats. But he said there are enough red flags to raise concern.

“Overall, I would say this is definitely suspicious,” he said.

Disaster relief fraud has perhaps existed as long as disaster relief itself. But with emerging technology like generative AI, the speed, scale, and sophistication with which disaster relief scams can be executed present a greater risk to consumers. 

Climate change and the greater frequency of disasters will likely result in more opportunities to harm vulnerable communities, especially those in disaster-prone regions like the Caribbean. With underreporting obscuring the scope of the problem and resources for enforcement stretched thin, there’s a widening gap between the sophistication of the scams and the ability to detect and stop them.

Whether Project PR Relief deliberately sought to trick well-meaning donors or was just an earnest but poorly executed effort to raise money for an important cause, the case illustrates that anyone with access to the internet and a few hours to spare can create an online presence that could mislead and potentially exploit both disaster victims and the people trying to help them. 

Puerto Rico, a commonwealth, has a contentious history with the United States. Residents there effectively have second-class citizenship. Gonzalez believes Project PR Relief is a scam, yet another affront to the people of Puerto Rico.

More Disasters, More Fraud

Hurricane Katrina slammed into the Gulf Coast on Aug. 29, 2005. The storm wreaked havoc across the southeastern United States, but low-income Black communities bore the brunt of the destruction.

Katrina, which remains one of the deadliest and costliest storms in U.S. history, reshaped how the country understands and responds to disasters. It also opened the door to an extraordinary amount of fraud.

Haywood Talcove, CEO of LexisNexis Risk Solutions for Government, which sells fraud detection and prevention tools to governments,# remembers how emergency aid meant for Katrina households was abused, including by people who fraudulently posed as victims. 

Talcove suspected then that, as the U.S. transitioned to a digital landscape, it would become even easier for schemers to take advantage of opportunities meant to help victims. 

“If someone’s hungry or someone’s been displaced because of a disaster, you want to make sure they have a roof over their head,” he said. “You want to make sure that they have food, you want to make sure they’re taken care of. And what’s happened is that empathy has allowed criminals to get into these programs."

Talcove’s prediction proved prescient: as fraud has moved online – including on social media and crowdfunding platforms, and now with help from AI – it has grown in scale and become more costly. In 2025, consumers lost $2.1 billion to scams that reached them on social media, according to the FTC. 

In 2019, a California woman pleaded guilty to theft and fraud for using social media to solicit thousands of dollars in donations of cash and supplies – ostensibly for firefighters battling California’s Holy Fire – and keeping them herself. A Texas man pleaded guilty in 2022 to defrauding the Red Cross of more than $46,000 meant for victims of Hurricane Harvey. And a woman who created a GoFundMe fundraiser for a friend who had lost her home in a California wildfire last year learned that an impostor had quickly made a nearly identical copy of her GoFundMe and was using it and a fake Instagram account to solicit donations. 

Further compounding the issue is that large-scale, billion-dollar disasters are becoming more frequent. Climate Central estimates that, so far this year, the cost of weather and climate disasters in the United States has surpassed $50 billion.

And as the frequency and scale of disasters increase, it’s likely that disaster assistance scams will also, according to a U.S. Government Accountability Office report issued in May. 

“When you increase things like the scope of people affected and the dollars involved in response and recovery, that makes it even more attractive to people who are interested in gaining financially through that crisis,” according to Rebecca Shea, the director of the GAO’s Forensic Audits and Investigative Service team. “It brings out an increased risk for fraud and an increased risk for scams against the survivors.”

AI-Assisted Scams

In an aerial photo posted to the Project PR Relief website, gray storm clouds roll in over colorful casitas dotting the coastline of Vieques, one of Puerto Rico’s smaller islands. It looks real enough, but a review by The Margin revealed the image contains an embedded AI watermark. 

Another image posted to the group’s social media depicts a volunteer helping an elderly man haul a five-gallon water jug up the side of a mountain, both awkwardly gripping the jug in a way that defies common sense.

Though Project PR Relief claims to have been established in 2017, its domain was registered in June, The Margin found. An earlier version of the site bore a tag for Lovable, a platform that promises users they can “build software in minutes.” It’s also been recognized as popular with scammers.

The site, marked up

The below is a rebuilt copy of the projectprrelief.org website combining current live text (as of October 26, 2026) and previously provided web pages from the Internet Archive, with sourced dates noted. Any provided text is copied word for word from the pages on the dates shown, with some passages cut (marked …). The layout has been simplified, with non-functional buttons and forms. Pages marked † have since been removed and come from the Internet Archive.

Beyond the website

These records are not part of the website and are rebuilt from screenshots, obtained by The Margin, of a donation email, an IRS charity search, and the group's Threads and Instagram accounts. The provided text is copied word for word except where a tab says it is a paraphrase. Any contact details have been removed.

Experts warn that artificial intelligence will only further facilitate fraud. The Deloitte Center for Financial Services predicted in 2024 that U.S. losses from fraud enabled by generative AI could reach $40 billion by 2027.

After the Hill Country, Texas floods in July 2025, a San Antonio TV news station found dozens of Facebook pages posting external donation links, purporting to raise money for victims. But those pages were filled with AI-generated images, including of celebrities like retired NFL quarterback Tom Brady rescuing a dog and singer George Strait comforting a man while waist-deep in flood water. According to the report, some of those Facebook pages had originated overseas. 

In California, the state attorney general’s office investigated fake sites that copied the state’s emergency disaster site and used an AI-generated image of the U.S. vice president. 

Last year, AI-generated images of Typhoon Halong, some linked to charities, made the rounds on Facebook, according to a local Alaska station. 

Shea of the GAO said technology like AI has allowed scams to become much more sophisticated.

“We’re all familiar with the Nigerian prince scam and these suggestions – ‘Look for spelling errors or sentence construction’” to identify bad actors, she said. “Well, not anymore. You just put it into ChatGPT and it puts out a perfect email for you… So all of the flags we've been told to look out for may not exist."

Shuttering the National Center for Disaster Fraud 

Shortly after Katrina made landfall in Louisiana, a Florida man named Gary S. Kraser set up the website www.AirKatrina.com. Air Katrina was, Kraser said, a group of pilots mobilizing to bring medical supplies to Louisiana and evacuate children and others in need.  

Kraser shared distressing details about what he said he witnessed, including desperate people flagging down help from their rooftops and dogs electrocuted after becoming tangled in electrical lines.  

On his website, he said he helped evacuate an infant who needed critical medical care: "I'm flying full loads out, and offering the empty plane on the outbound for Air Ambulance. If we didn't have the plane, I don't think the little baby would have survived... She is undergoing transplant surgery at this moment I am writing... 7 mos old, and smiling the whole way, as if she knew..."

During a phone call with one donor, Kraser, who said he was in the cockpit of a plane, was interrupted by apparent noise from air traffic control, according to reporting from the Tampa Bay Times. He told the woman on the call that without more money, that flight would be his last.

But Kraser was not really in a plane; the noises had come from his computer. He had not flown to Louisiana. And he wasn’t a licensed pilot. Yet he managed to collect nearly $40,000 from 48 donors who had tried to support his purported mission by sending money to his PayPal account or wiring it directly into his bank account.

Kraser’s case would become the first federal prosecution involving a fraudulent charity website tied to Hurricane Katrina. It was also an early win for the Hurricane Katrina Fraud Task Force, a collaborative effort among the Department of Justice and a number of federal and state agencies to more efficiently weed out the fraud that had become rampant in the wake of the storm. 

A September 2005 op-ed in the Las Vegas Sun, titled “The worst of the worst,” cited a Missouri case in which a group of sites soliciting donations for Katrina was directing users to donate via another site with ties to white supremacist groups.

Days later, government officials announced that about 4,000 websites advertising Hurricane Katrina-related relief services had already popped up on the internet. Officials found that, of the 2,100 sites they had reviewed at that point, sixty percent had originated overseas.

By February 2006, more than 200 people had been charged with various hurricane-fraud-related crimes. The task force had evolved, by 2008, into the National Center for Disaster Fraud and was authorized to collect complaints related to any type of disaster. At its nexus was a call center housed at Louisiana State University responsible for filtering through reports and routing them to the appropriate federal, state, or local agency.

The NCDF was a more efficient way to handle complaints from across the country, and it collected nuanced data about the types of scams proliferating.

This 2018 Department of Justice PSA promotes the National Center for Disaster Fraud. Until March 31, 2026, the NCDF tracked complains of fraudulent activity and disaster scams.

In a 2018 interview with Homeland Security Today, then-U.S. Attorney for the Middle District of Louisiana Brandon Fremin, who also served as the executive director of the NCDF, said the agency could help spot national trends to identify offenders operating across jurisdictions.

“If those victims of fraudulent activity perpetrated by that particular contractor are notifying the local authorities and many jurisdictions, then you may be left with potentially several smaller cases,” Fremin said. “If every one of those complaints has been run through the NCDF, then we would notice a trend, and we could then perhaps build a larger-scale, and hopefully stronger, case across jurisdictional lines and then, obviously, send a stronger message. For me that’s one of the most important functions of the NCDF.”

In more than 20 years, the NCDF collected more than a million complaints. But this past March, the Department of Justice announced that it had shuttered the center, after a program review determined that the “NCDF’s intake function was no longer the most effective avenue for pursuing disaster fraud.” The closure is expected to save the agency $600,000 annually. 

The DOJ site now lists nine different options for reporting disaster fraud. The agency did not respond to multiple requests for comment. 

The true prevalence of disaster scams is difficult to quantify because of underreporting.  

Even so, in August 2024, the U.S. Census estimated that between 1.9 and 2.7 million, or 61.5% to 77.3%, of displaced people had experienced a possible scam offer within one month of experiencing a disaster. A June 2025 Harris Poll found that more than a third of Americans have experienced “fraudulent activities,” such as identity theft, government assistance fraud, and charity fraud, after a disaster.

The FBI’s Internet Crime Complaint Center received more than 4,500 complaints reporting around $96 million in losses to fraudulent charities, crowdfunding accounts, and disaster relief campaigns in 2024 alone.  

And many fictitious websites disappear before they can be tracked, making it difficult to prosecute the people behind them. That so many of them are of international provenance further impedes prosecution.

The ‘Sad Thing’ About ‘Sham Charity’  

Project PR Relief might be a fake charity, but the transactions were real. Individuals who donated to the site received an automated email thanking them for their donation and a second one with a receipt.  

In August, in response to the criticism on social media, the organization posted on Threads that its nonprofit status was being rectified.  

“We are a legitimate organization working to support relief efforts in Puerto Rico,” the post said. “We obtained our EIN and hired someone to properly complete the nonprofit process, but unfortunately that process was not completed as it should have been.”

Project PR Relief had been using the payment processor Stripe to accept money. Once concerns began to circulate online, the actors behind the site quickly paused donations. 

Requests for comment sent to email addresses associated with the website could not be delivered. A phone number that had been listed on the website was disconnected. And a message The Margin sent via Instagram went unanswered.

After a user reported Project PR Relief to Linktree, a tool that lets users list multiple links in their social media bios, Linktree said it took “enforcement action against the reported account due to violations involving deceptive donation practices and suspected financial fraud.” 

At least one donor received a refund from Project PR Relief, but there have been no further updates from the organization. The site, however, ranks high in Google’s search results for queries about how to help with Puerto Rico’s water crisis. "We design our ranking systems to surface high-quality sites and to prevent spam and scams from appearing in our results," Google said in a statement. "We’re constantly updating our systems and enforcing our policies to stay ahead of spammers and bad actors."

Consumer advocates at the National Consumer Law Center see holding payment processors accountable for fraud as key to protecting consumers, because processors can enable fraud. They make a profit with each transaction, and they keep processing fees even if a transaction is disputed.  

There’s an added risk for people who use their bank accounts to make the donations because that offers fewer protections than credit cards. On Threads, some users said they were having difficulty getting their banks to dispute the charges.  

“If you actually authorize the transaction – so you hit send – then you can't complain to your bank that it was unauthorized,” said Carla Sanchez-Adams, a senior attorney at the National Consumer Law Center. “There is a huge gap in the law right now as to what happens when you're defrauded by a scam versus when someone does an account takeover and takes your money without you consenting to it and hitting send.”  

Stripe declined to comment on the Project PR Relief website. In response to follow-up requests for comment, the company referred to its policies on unacceptable risk and prohibited and restricted businesses. 

According to the company’s policy, Stripe uses “a combination of machine learning models, heuristics and human reviews to detect” bad actors. 

Dotan Weinman, an assistant director of the Bureau of Consumer Protection at the Federal Trade Commission, said underreporting makes it harder to ferret out fraud, particularly in sham charity cases, because consumers may not realize the organization is fake.  

“When we get a product or we get a service, and we don’t get what was advertised, then we learn that we were duped and we can do something about it,” he said. “In a charity, you might get a letter, a thank you for the donation, but you have no idea where the money actually went.” 

Some of the FTC’s enforcement actions against sham charities have targeted organizations that claimed to raise money for important causes but used only a small fraction of donations for their intended purpose or engaged in deceptive fundraising practices.

Sanchez-Adams said that payment processors should take a preventative approach to protecting consumers before fraud happens. One example is “verification of payee,”  which lets users confirm the name of the person or business receiving the money. 

“Why do you have to wait until after (there are) problems?” she said. “You should be doing more on the front-end to screen to make sure that bad actors aren't getting on the platform to begin with."

‘We’re not getting help from anywhere else’ 

In October, as Hurricane Isaias advanced across the Gulf of Mexico toward Florida’s panhandle, artificial intelligence models trained on historical weather data were lauded for their ability to predict where the storm would make landfall.  

At the same time, watchdog groups were advising consumers to watch for AI-fueled charity scams that could emerge after the storm. 

AI has positive uses, including helping catch bad actors, Acharya said, but it can also widen the potential pool of scammers and make those bad actors more efficient. 

“The rapid progress we are having is amazing, but the policing being done is not at the efficiency we want to be at,” Acharya said. “That's what scares me as a researcher too because it can have both impacts.” 

While there are efforts at the federal level to address the threat of consumer scams, even the GAO contends that the government could do more to protect consumers. 

Preemptive communications may help. In Hawaii, an initiative called Help Maui Rise compiled a list of vetted GoFundMe fundraisers after wildfires and storms and shared it on social media and through local news outlets.

Ultimately, much of the existing guidance directs consumers to educate themselves.   

Sukrit Venkatagiri, an assistant professor at Swarthmore College who studies misinformation and the use of AI, said that approach allows fraudsters to operate in a vacuum where the government puts the onus of identifying what’s real and what’s not on the individual. 

He and his team are studying how web pages of financial institutions, peer-to-peer payment platforms, and social media companies talk about scams and support their customers. Initial findings show that much of the language around scams is framed to advise consumers to better protect themselves, he said.  

Venkatagiri said opportunities for redress, such as directing consumers to contact their bank or to submit a report to a government agency, are minimal. And even then, government agencies won’t have the resources to pursue every complaint it receives.  

“They’re going to focus on overarching patterns, rather than individual cases,” he said. “So what happens to the individual? What are they left to do?” 

Gonzalez said that in Puerto Rico, where residents already contend with dilapidated infrastructure, a high cost of living, a failing electrical grid, and steep utility bills, the threat of scammers only further disempowers people. 

“Take away water issues. Take away power issues. There’s just so many things that people are down bad about already on the island,” she said. “And so to have people scamming people that are genuinely trying to help – we’re not getting help from anywhere else."

***

Your donation supports in-depth reporting, data analysis, data visualizations, and original photography. Join us in the fight for environmental justice by contributing today — every dollar helps us tell the stories of the people and communities who have been and are currently experiencing injustice.