Researchers warn of increasingly deceptive artificial intelligence scams

Computer software can now scan public social media accounts and construct remarkably believable fake emails within seconds. According to a new study from researchers at the University of Texas at Arlington, online scammers no longer need to write messages by hand to target individuals. Instead, automated tools can read recent posts about vacations, concerts, or daily routines, using those intimate details to fabricate believable messages designed to trick people into revealing passwords or payment details.

This shift makes tailored scams cheap and easy to run on a massive scale. In testing, the researchers found that built in security filters for several prominent artificial intelligence tools failed to block malicious prompts between 81 and 100 percent of the time. The safety systems only succeeded consistently after receiving explicit instructions on how to recognize tailored fraud, which raised detection rates up to 99 percent.

While software providers can patch their tools, it remains unclear how quickly companies will adopt these stronger protections across all public models. A major open question is whether social media networks will take stronger action to restrict automated web scrapers from harvesting personal posts in the first place. Until both sides act, sharing routine life updates online could carry an unprecedented security risk.