In a world increasingly flooded with content generated by artificial intelligence, big tech companies are now embroiled in a high-stakes battle to develop cutting-edge tools to tell the real from the fake.
from OpenAILaunch of new system to identify image created independently Darui 3 text to image generator AmazonIntroducing AI to spot fraudulent reviews, businesses are trying to stay one step ahead of malicious actors looking to exploit their platforms. As AI continues to proliferate across industries, developing robust fake detection tools has become a priority, with far-reaching implications for fields from e-commerce to journalism to politics.
“Retailers can leverage AI to more effectively combat fraud, encounter bad actors in the field, and even beat them at their own game.” sophia carltonFraud Reform Executive Accenturehe told PYMNTS.
PYMNTS Intelligence’s “Managing fraud in online transactions”Last year, 82% of Americans e-commerce Merchants selling internationally faced a cyber breach, with nearly half losing customers and revenue, and 68% struggling to balance security and customer satisfaction.
AI tools to fight fraud
A successful fraud attack can negatively impact a retailer's reputation. Christoph van de Weijr, CEO of an identity solutions company telesignhe told PYMNTS.For example: Telesign Annual Confidence Indexalmost half of fraud victims investigated They blame the affected brand, with 64% reporting a negative impact on their perception of the brand, an increase of 36% from last year.
“This negative perception can spread quickly as victims share their experiences,” he says. “37% advised friends and family to avoid the brand, and 34% posted about the fraud incident on social media, which could reach a wide audience around the world. It highlights the need for solutions. both Customer identity and revenue sources. By leveraging AI fraud prevention solutions, organizations can detect and respond to fraud in real-time, reducing financial losses and maintaining customer trust in an increasingly competitive retail environment. ”
Companies focus on growth scam There is a problem and we are working to prevent it. OpenAI announced on Wednesday (May 7): Introduction to tools Designed to identify content generated by the DALL-E 3 text-to-image generator. Additionally, the company is now accepting applications for the first group of testers for its image detection classifier.tool designed Evaluate the probability that the image was created by DALL-E 3.
“Our goal is to evaluate the effectiveness of the classifier, analyze its real-world applications, identify considerations related to such use, and independently investigate the characteristics of content generated by AI. OpenAI said: in a statement.
Machine learning typically does not take immediate action when it detects suspicious behavior. Keegan KeplingerSenior Threat Researcher at a Security Company e-centaia he told PYMNTS. Instead, we use specialized models to score activity based on how unusual it is and whether it matches known fraud patterns. This scoring helps human analysts decide which cases to investigate first based on the level of risk. Machine learning algorithms analyze various aspects of each transaction, including time, location, amount, and parties involved. These parties may have a transaction history that suggests fraudulent activity.
“A common example is when your credit card company blocks your purchase or calls you after you try to make a transaction while you're traveling,” he says. “Your spending habits before your trip establish a baseline for your transaction history, but sudden changes in location and sometimes large transaction amounts associated with your trip could be signs of credit card fraud. For example, they can be skimmed from e-commerce sites by hackers, sold, and used by criminals.
OpenAI says the tool accurately detected approximately 98% of images created by DALL-E 3 and incorrectly identified less than 0.5% of artificial images as being generated by AI. The company further noted that common image modifications such as compression, cropping, and saturation changes have minimal impact on the tool's effectiveness. However, he acknowledged that other modifications may impair performance. The classifier also showed reduced effectiveness in distinguishing between images generated by DALL-E 3 and images generated from other AI models.
Carlton explained that generative AI can help generate new programming code, complete partially written code, and convert code between programming languages. These applications could lead to “more effective fraud models, faster model development for new schemes, or more efficient fraud model coordination and management,” she said.
A.I. It is set Revolutionizing retailer fraud prevention and detection in the futuresaid Carlton.. By leveraging its capabilities, retailers can better manage fraud, reduce losses and costs, and better protect their customers.
“A.I. is placed Carlton said it will revolutionize fraud prevention and detection for retailers, adding that it will enable retailers to deal with fraud “more efficiently and effectively than ever before.”
Amazon’s fraud prevention efforts
Amazon is also ramping up its use of AI to combat fraudulent reviews. The company reported that its AI system has blocked over 200 million suspects. fake reviews Around the world in 2022. this This is part of Amazon's strategy to maintain the integrity of its review system, which benefits consumers who rely on these ratings to make purchasing decisions and businesses who rely on genuine customer feedback. It's essential.
“Fake reviews intentionally mislead customers by providing information that is not fair, authentic, or not targeted at the product or service.” Josh Meeksaid a senior data science manager on Amazon's fraud prevention team in an April blog post. “Not only do millions of customers trust the credibility of Amazon reviews when making purchasing decisions, but millions of brands and businesses can accurately identify fake reviews and ensure that their customers receive them.” We ensure that our reviews reflect the opinions of real customers and protect honest sellers who rely on us for accurate ratings. We strive to monitor and enforce our policies responsibly.”
Andrew SellersHead of Technology Strategy Confluencetold PYMNTS, “AI can build highly detailed models to assess fraud risk that incorporate many characteristics of end-customer behavior, such as transaction time, amount, location, purchase/billing history, etc.” he said. These models are based on rules defined by experts or Machine learned patterns From transaction data.
“Running this type of analysis at scale requires automation,” Sellers added.
Then, To prevent fraud in the retail sector, Sellers says, “AI models can assess the risk of fraud in real-time as point-of-sale transactions occur. If so, this fraud risk characterization may occur along the purchase approval process. This real-time assessment ensures highly suspicious transactions are prevented. is rejected Even if the account is displayed things to do Excellent. “
Looking to the future, Sellers predicted that “AI algorithms will continue to be more accurate in their evaluations and more precise in taking into account a consumer's individual circumstances.”
For all of our coverage of PYMNTS AI, subscribe to our daily subscription AI Newsletter.
