AI-Powered Scams Surge as Fraudsters Manipulate LLMs and Automate Deception
A new report from F-Secure reveals that cybercriminals are increasingly weaponizing artificial intelligence to poison search results and automate the creation of highly convincing, deceptive content.
The landscape of digital fraud is undergoing a significant transformation as cybercriminals increasingly leverage artificial intelligence to enhance the efficacy and reach of their campaigns. According to the *F-Secure Scam Intelligence & Impacts Report 2026*, while the overall volume of scam attempts has remained stable over the past year, the sophistication of these attacks is rising. More than half of consumers now encounter scam attempts at least monthly, with the United States reporting the highest levels of exposure among surveyed markets Help Net Security.
The technical mechanism behind this shift involves the exploitation of AI systems through data poisoning and the manipulation of Large Language Model (LLM) outputs. As users increasingly rely on AI assistants as primary search tools, attackers are finding new avenues to inject fraudulent information. In internal testing conducted by F-Secure, researchers discovered that ChatGPT provided fraudulent airline customer service numbers during a major U.S. winter storm that resulted in over 5,000 flight cancellations. This proof-of-concept demonstrates that attackers can manipulate AI responses to intercept users seeking legitimate technical or customer support Help Net Security.
Beyond search manipulation, AI is being integrated into the shopping experience to deceive consumers. AI-powered shopping tools can inadvertently validate fraudulent merchants, making them appear trustworthy at the critical point of purchase. In one test, ChatGPT repeated a fake store’s claim that it was "going out of business," effectively amplifying the urgency messaging used by the fraudulent merchant. Furthermore, 89% of observed AI use by scammers is focused on improving the quality of "scam bait," including the production of polished messages, convincing visuals, and synthetic voices that are increasingly difficult for users to identify Help Net Security.
The impact of these evolving tactics is tangible, with more than half of scam victims reporting monetary loss in 2026—a figure that has more than doubled since 2025. While younger consumers report higher overall scam activity due to their heavy digital presence, older victims, particularly those aged 65 to 74, are more likely to suffer financial loss when targeted. Fraud schemes involving direct financial transfers, such as fake invoices, debt schemes, and investment fraud, currently account for half of all reported attempts, with fake invoice scams seeing the largest year-over-year increase Help Net Security.
As these threats evolve, public anxiety is mounting, with 84% of respondents expressing concern that AI will make it impossible to distinguish between genuine and fraudulent content online. The research highlights that individual vulnerability is deeply tied to digital maturity and institutional trust. Consequently, consumers are increasingly viewing cybersecurity as a critical service feature, with many indicating a willingness to switch providers based on the strength of available security protections Help Net Security.
This trend underscores a broader shift in the scam economy, where the barrier to entry for high-quality fraud is being lowered by automation. As fraudsters continue to experiment with the underlying data and logic of AI assistants, the focus for both security researchers and consumers must shift toward verifying the integrity of AI-generated information and maintaining skepticism toward automated recommendations. Future developments will likely see continued efforts to weaponize LLMs, necessitating more robust defenses against data poisoning and prompt manipulation Help Net Security.