The GPT-4 logo is seen in this photo illustration on 13 March, 2023 in Warsaw, Poland.
Jaap Arriens | Nurphoto | Getty Images
AMSTERDAM, Netherlands — Major banks and fintech companies claim to be piling into generative artificial intelligence as the hype surrounding the buzzy technology shows no signs of fizzling out — but there are lingering fears about potential pitfalls and risks.
At the Money 20/20 fintech conference in Amsterdam, Netherlands, executives at large lenders and online finance firms sang the praises of generative AI, calling it an “explosion of innovation,” and saying it will “unleash innovation in areas that we can’t even think about.”
related investing news
9 hours ago
Chalapathy Neti, head of AI at global bank messaging network Swift, described the progress made with ChatGPT and GPT-4 as “mind-boggling.” He added, “This is truly a transformative moment.”
But in the short term, banks are scrambling to figure out the use cases.
The Netherlands’ ABN Amro is one banking giant that’s piloting the use of generative AI in its processes.
Annerie Vreugdenhil, chief commercial officer of ABN Amro’s personal and business banking division, revealed on a panel that it is using the technology to automatically summarize conversations between bank staff and customers. It’s also using it to help its employees gather data on customers to assist with answering queries and avoid repetitive questions.
The bank is now in the process of scaling these pilots to 200 employees and is exploring a number of new pilots to start this summer.
In a closed-door session on the application of AI in financial services, meanwhile, two banking executives explained how they’re using the technology to improve their internal code and analyze how their clients are behaving.
“We are experimenting at this stage and we don’t have necessarily anything client facing but we are using the [tech the] same as other companies, for example, code refactoring, comms calls, the other way around,” said Mariana Gomez de la Villa, an executive at ING Bank specializing in strategy and innovation.
Indeed, the banks appeared unanimous in their hesitation to roll out ChatGPT-like tools to customer-facing scenarios.
Jon Ander Beracoechea Alava, advanced analytics discipline head at Spanish bank BBVA, said that the lender had taken a “conservative approach” to AI, adding that, at this stage, generative AI is “still early” and “immature.”
A crucial issue is that advanced AI systems require the processing of huge volumes of data — a sensitive commodity wrapped up in all kinds of rules and regulations. As such, Alava said that at this stage it was too “risky” to involve sensitive information from customers.
Generative A.I., explained
Generative AI is a specific form of AI that is able to produce content from scratch. The systems take inputs from the user and feed them into powerful algorithms fueled by large datasets to generate new text, images and video in a way that’s more humanlike than most AI tools already on the market.
The technology was thrust into the spotlight following the success of OpenAI’s GPT language processing technology. ChatGPT, which uses massive language models to create human-sounding responses to questions, has ignited an arms race among some companies over what is seen as the next “paradigm shift” in tech.
In March, Goldman Sachs‘ chief information officer, Marco Argenti, told CNBC the bank is experimenting with generative AI tools internally to help its developers automatically generate and test code.
More recently, in May, Goldman spun off the first startup from the bank’s internal incubator — an AI-powered social media company for corporate use called Louisa. The push into AI is part of a larger effort by CEO David Solomon to expedite the bank’s digital makeover.
Morgan Stanley, meanwhile, is using it to inform its financial advisors on queries they may have. The bank has been testing an OpenAI-powered chatbot with 300 advisors so far, with a view to ultimately aid its roughly 16,000 advisors in making use of Morgan Stanley’s repository of research and data, according to Jeff McMillan, head of analytics and data at the firm’s wealth management division.
A.I. ‘co-pilot’
These are just some examples of how financial firms are using AI, but more as a digital helper than as a core part of their services.
Gudmundur Kristjansson, CEO and co-founder of Icelandic regulatory technology firm Lucinity, showed CNBC how artificial intelligence can be used to assist with a key area in finance: fighting crime.
An AI tool the company created, called Luci, aims to help compliance professionals with their investigations. In a live demonstration, Kristjansson showed himself looking into a money laundering case. The AI tool analyzed the case and described what it saw and then completed an independent review.
In this use case, the AI acts as more of a resource — or “copilot” — to help an employee find data and flesh out a case rather than replace the role of a person looking into reports of suspicious activity.
“Where you find money laundering is through … interconnected networks of people who are basically employed to do it. That’s why it’s so hard to find it. Banks spent this year $274 billion on prevention,” Kristjansson told CNBC in an interview.
He said where Luci helps is by vastly reducing the amount of time spent trying to work out whether something is fraud or money laundering.
The whole appeal of AI to the big banks and fintechs, Money 20/20 attendees said, is the potential reduction in the time and money it takes to complete tasks that can take human employees days.
Niklas Guske, chief operating officer at Taktile, a startup that helps fintechs automate decision-making, acknowledged that the use of AI is challenging in the financial sector, given the lack of publicly available data.
But he stressed that it could be a “crucial” tool to reduce the companies’ operational expenses and improve efficiency.
“In many fintech applications, this is done through an increase in automation and reducing manual processes, especially in onboarding and underwriting,” he told CNBC.
“This automation is truly enabled through access to more data sources, which empower lenders to gain new insights and identify the right customers without having to parse through dozens of PDFs for the right piece of information.”
Returns on Amazon are free and easy for shoppers, but they’re risky and expensive for the small businesses that sell a majority of the goods on the world’s biggest e-commerce site. Returns have driven some sellers to exit the popular Fulfillment by Amazon program, while others told CNBC they’d like to leave the platform altogether.
At the heart of the problem is a big rise in returns fraud, which has led to customers mistakenly receiving used products when they ordered something new. In two particularly egregious examples involving baby products described to CNBC, Amazon sent customers used diapers and a chiller with someone else’s rotten breastmilk inside.
“I really don’t think that consumers understand how many small businesses are on Amazon and how their return habits affect small businesses and families like mine,” said Rachelle Baron, owner of Beau and Belle Littles, which sells reusable swim diapers on Amazon.
Baron said her business tanked after a return incident with Amazon. The e-commerce platform shipped soiled swim diapers to customers after the used baby products had been returned to Amazon, Baron said.
“There was actually two diapers that were sent out that were poopy,” she said.
In 2024, nearly 14% of all U.S. retailreturns were fraudulent, up from 5% in 2018, according to a report by the National Retail Federation. In total, the report found that returns cost retailers $890 billion in 2024.
Amazon started charging sellers in its fulfillment program (FBA) a new fee in June 2024 for items that exceed certain return rate thresholds. Sellers who sign up for FBA rely on Amazon for logistics, including shipping, packing and returns.
In September, a couple months after the fee went into effect, e-commerce group Helium 10 saw return rates for U.S. Amazon sellers drop nearly 5%.
“It’s forcing the seller to have higher quality listings and higher quality products,” said Helium 10 General Manager Zoe Lu.
Amazon has alsostarted adding a warning label to some “frequently returned items,” which could be contributing to the dip.
Rising prices
However, the new fee may also be leading to rising prices.
One survey by e-commerce analysis company SmartScout found that 65% of sellers said they raised prices in 2024 directly because of Amazon fee changes. Other sellers told CNBC returns fraud is the reason they’ve raised prices.
In total, CNBC talked to seven Amazon sellers to find out how they’re handling the rising cost of returns.
“We’re running at about just over 1% net profit on Amazon, totally due to fraud and return abuse,” said Lorie Corlett, who sells Sterling Spectrum protective cases for hot wheels. She said her return rate is 4% on Amazon and only 1% on other marketplaces like Walmart. “It’s really Amazon that’s accountable at the end of the day. People would stop doing it if Amazon held them accountable.”
Amazon told CNBC it has no tolerance for fraudulent returns and that it takes action against some scammers. Those measures include denying refunds and requiring customer identity verification.
Mike Jelliff sells professional music gear through his GeekStands brand on Amazon and eight other marketplaces. He said his return rate on Amazon is three times higher than the average he sees elsewhere.
“On eBay, we’re allowed to block specific customers out,” Jelliff said. “But on Amazon, that customer is still allowed to repurchase from us.”
Jelliff showed CNBC the system of about 40 cameras he’s installed in his Tyler, Texas, warehouse to track every outgoing item, incoming return and unboxing. He uses the images when filing appeals with Amazon, including when customers request refunds claiming they never receive an item. He keeps a blacklist of repeat offenders who commit this kind of fraud and those who return used and damaged items, which become a total loss for him.
Amazon has made some improvements to its returns process, said Jelliff, who doesn’t rely on FBA. This includes Amazon allowing small businesses to make multiple appeals when fighting a fraudulent return. Amazon has also let Jelliff opt-out of automatic return labels for items above $100 starting in 2023, and his return rate has been dropping since.
Mike Jelliff at his GeekStands warehouse in Tyler, Texas, on June 6, 2025. Jelliff sees three times more returns of his professional music gear on Amazon, compared to the average on other marketplaces like eBay and Walmart.
Jacob Schatz
Why returns are destroyed
Figuring out which returns are fraudulent and which are ready for re-sale is labor-intensive and item specific, experts said. That creates plenty of room for error.
“Because it’s such a large operation, things are missed,” said Lu of Helium 10. “I think they’re probably missed on the margins, but these stories are very impactful because it is such a reckoning for the brand.”
Ceres Chill founder Lisa Myers, who once relied on Amazon to handle returns for her business as part of FBA, has one of these stories.
In 2023, Amazon sent one of Ceres Chill’s products to a customer with someone else’s rotten breastmilk inside, said Myers, adding that the customer wrote a review saying, “she will never forget that smell.”
“To have something, and I don’t mean to be dramatic, but dangerous, somebody else’s bodily fluids in your kitchen rotting in something that you had intended to use for your child is unacceptable,” Myers said. “That’s the moment I broke down crying and just sat down and thought, I have no idea how this could have happened.”
Myers said she left FBA after the incident, leaving behind benefits like having her products labeled with Amazon’s Prime badge.
“It hurts our business to not participate in Fulfilled by Amazon,” Myers said.“It’s just we’re not willing to, we will never put profit over the safety and, frankly, mental health of our customers.”
Instead, Myers outsources all her returns to baby resell specialist Goodbuy Gear, which is on track to re-sell 200,000 returned baby products this year.
Re-selling responsibly
Kristin Langenfeld started GoodBuy Gear when she was a new mom struggling to find a good quality, used jogging stroller.
“We’ve spent the last nine years building out a database that has all of the products and the variations, the common issues, the recalls,” Langenfeld said. “For some of these, there’s 40 points that we inspect on the item itself, and it’s really complicated.”
Langenfeld showed CNBC the process at her warehouse in Malvern, Pennsylvania, where each item is inspected for about 15 minutes and is typically handled by at least four employees. The resource intensive process is paying off. She says 33 new sellers signed up in 2024, three times more than the previous year. And with business growing 50% year-over-year, she’s upgrading to a bigger warehouse in Columbus, Ohio.
She was inspired to handle returns after visiting a major retailer’s returns warehouse five years ago.
“Taped on the floor were signs that said ‘incinerate,’ ‘destroy,'” she said.
Returns generated an estimated 29 million metric tons of carbon emissions in 2024, and 9.8 billion pounds of returns ended up in landfills, according to reverse logistics software provider Optoro.
Amazon has faced criticism for destroying millions of pounds of unused products. In 2022, Amazon told CNBC it was “working towards a goal of zero product disposal,” but wouldn’t give a timeline for that ambition. Three years later, that goal is still in the works, with Amazon telling CNBC in a statement, “The vast majority of returns are resold as new or used, returned to selling partners, liquidated, or donated.”
In 2020, Amazon added two new options for sellers to re-home returns. “Grade and Resell” allows all U.S. FBA sellers to have Amazon rate the return and mark it as “used” before re-selling it. FBA Liquidation allows sellers to recoup some losses by offloading palettes of goods for re-sale on the secondary market through liquidation partners like Liquidity Services.
There’s also an FBA Donations program that’s been around since 2019, allowing sellers to automatically offer eligible overstock and returns to charity groups through the non-profit Good360. Amazon told CNBC these seller programs give a second life to more than 300 million items a year.
For shoppers wanting to keep returns from incineration or landfills, Amazon also has options.
Amazon Resale has used and open-box goods, Amazon Renewed sells refurbished items and Amazon Outlet sells overstock. Daily deal site Woot!, bought by Amazon for $110 million in 2010, also sells scratched and dented items. Customers can also trade in certain electronics, like Amazon devices, phones and tablets, for Amazon gift cards or send them to the company’s certified recycler.
“I hope the change that we’re able to make as a country is that we stop making crap,” Langenfeld said. “We should make high quality products that are meant for resale.”
Meta approached artificial intelligence startup Perplexity AI about a potential takeover bid before ultimately investing $14.3 billion into Scale AI, CNBC confirmed on Friday.
The two companies did not finalize a deal, according to two people familiar with the matter who asked not to be named because of the confidential nature of the negotiations.
One person familiar with the talks said it was “mutually dissolved,” while another person familiar with the matter said Perplexity walked away from a potential deal.
Bloomberg earlier reported the talks between Meta and Perplexity. Perplexity declined to comment. Meta did not immediately respond to CNBC’s request for comment.
Meta’s attempt to purchase Perplexity serves as the latest example of Mark Zuckerberg‘s aggressive push to bolster his company’s AI efforts amid fierce competition from OpenAI and Google parent Alphabet. Zuckerberg has grown agitated that rivals like OpenAI appear to be ahead in both underlying AI models and consumer-facing apps, and he is going to extreme lengths to hire top AI talent, as CNBC has previously reported.
Read more CNBC reporting on AI
Meta now has a 49% stake in Scale after its multibillion-dollar investment, though the social media company will not have any voting power. Scale AI’s founder Alexandr Wang, along with a small number of other Scale employees, will join Meta as part of the agreement.
Earlier this year, Meta also tried to acquire Safe Superintelligence, which was reportedly valued at $32 billion in a fundraising round in April, as CNBC reported on Thursday.
Daniel Gross, the CEO of Safe Superintelligence, and former GitHub CEO Nat Friedman are joining Meta’s AI efforts, where they will work on products under Wang. Gross runs a venture capital firm with Friedman called NFDG, their combined initials, and Meta will get a stake in the firm.
OpenAI CEO Sam Altman said on the latest episode of the “Uncapped” podcast, which is hosted by his brother, that Meta had tried to poach OpenAI employees by offering signing bonuses as high as $100 million with even larger annual compensation packages.
“I’ve heard that Meta thinks of us as their biggest competitor,” Altman said on the podcast. “Their current AI efforts have not worked as well as they have hoped and I respect being aggressive and continuing to try new things.”
Ether ETFs have finally come to life this year after some started to fear they may be becoming zombie funds.
Collectively, the funds tracking the price of spot ether are on pace for their sixth consecutive week of inflows and eight positive week in the last nine, according to SoSoValue.
“What we’re seeing is institutional recalibration,” said Ben Kurland, CEO at crypto charting and research platform DYOR. “After the initial ETH ETF approval fizzled without a price pop, smart money started quietly building positions. They’re betting not on price momentum but on positioning ahead of utility unlocks like staking access, options listings, and eventually inflows from retirement platforms.”
The first year of ether ETFs, which launched in July 2024, has been characterized by weak demand. While the funds have had spikes in inflows, they’ve trailed far behind bitcoin ETFs in both inflows and investor attention – amassing about $3.9 billion in net inflows since listing versus bitcoin ETFs’ $36 billion in their first year of trading.
“With increasing acceptance of crypto on Wall Street, especially now as a means for payments and remittances, investors are being drawn to ETH ETFs,” said Chris Rhine, head of liquid active strategies at Galaxy Digital.
Additionally, he added, the CME basis on ether – or the price difference between ether futures and the spot price – is higher than that of bitcoin, giving arbitrageurs an opportunity to profit by going long on ether ETFs while shorting futures (a common trading strategy) and contributing to the uptrend in ether ETF inflows.
Stock Chart IconStock chart icon
Ether (ETH) 1 month
Despite the uptrend in inflows, the price of ether itself is negative for this month and flat over the past month.
For the year, it’s down 25% as it’s been suffering from an identity crisis fueled by uncertainty about Ethereum’s value proposition, weaker revenue since its last big technical upgrade and increasing competition from Solana. Market volatility driven by geopolitical uncertainty this year has not helped.
In March, Standard Chartered slashed its ether price target by more than half. However, the firm also said the coin could still see a turnaround this year.
Since last week’s big spike in inflows, they’ve “slowed but stayed net positive, suggesting conviction, not hype,” Kurland said. “The market looks like a heart monitor, but the buyers are treating it like a long-term infrastructure bet.”
Don’t miss these cryptocurrency insights from CNBC Pro: