By Robert McMillan and Amrith Ramkumar
The scammers who have been bombarding American smartphones with fake messages about packages and toll-road fines have pivoted to impersonating the phone companies.
Worse, they are using artificial intelligence's coding superpowers to create fake websites that look like the real thing, according to Google, which on Friday sued a group of scammers it describes as one of the most prolific bad actors in the spamosphere. The company said it is the first case against a defendant employing Google's Gemini AI model.
Google and law enforcement call the group "Outsider" and say it is sending messages telling people they have mobile-phone-company rewards points to use up. To pressure victims into giving up their account information, they tell them they need to log in immediately before their points expire to claim such things as free headphones or Apple Watches.
A link in the message goes to a website modeled after the phone carrier. To make it easier to create such websites, the group developed a guide for using Gemini to generate computer code, one of the most popular uses of AI models. The ultimate goal of many of these scams is to steal credit card numbers that can then be resold or used to buy gift cards or luxury items.
The addition of AI allowed the group to circulate hundreds of website templates. It used the same tactics by telling consumers there was a problem with their brokerage accounts. Using more than 8,000 phishing websites, Outsider has stolen an estimated 3.87 million credit card numbers from victims in dozens of countries, leading to $1.9 billion in losses since July 2023, the Federal Bureau of Investigation said. The FBI and telecom companies are working with Google on its case.
"Criminals increasingly use AI to make fraud like this more convincing and harder to detect," said Brett Leatherman, assistant director of the FBI's cyber division.
Google says it received some 55,000 reports of suspicious messages on Google Messages, the default text platform for Android users, in the two-week period that ended June 1, including many allegedly from Outsider. Because of AI, the total number of websites that could be created in a scheme like Outsider's is effectively unlimited, Google said.
Telcom-focused scams like the ones perpetrated by Outsider have blown up over the past year, according to the threat-intelligence firm WMC Global. For example, phishing messages targeting one U.S. telecommunications company have increased 10-fold just over the past month.
Hackers and cybercriminals have used AI for a variety of schemes. Last year, Anthropic said that AI was being used for ransomware, to analyze stolen data and to create fake identities for fraud. The company also warned that state sponsored hackers from China were using AI to automate cyberattacks.
In addition to creating believable phishing messages, the technology has been used for podcasting. On Thursday, Sen. Maggie Hassan (D., N.H.) said music-streaming platform Spotify had removed content or banned creators tied to some 57,000 podcast episodes that included drug-related content after a Senate committee investigation, a challenge that was exacerbated by AI. A Spotify spokeswoman said the company polices spam aggressively and has been removing content related to this attack for more than a year.
Google has filters on Gemini that flag potentially illegal or dangerous interactions users have with the chatbot, but generic requests like the ones used by Outsider to generate website code wouldn't normally be flagged, Halimah DeLaine Prado, Google's general counsel, said in an interview. "A content filter isn't necessarily going to block a person's ability to create code for a website," she said. "That's a fairly innocuous ask."
Google is asking a New York federal court to halt the activity of the alleged cybercriminals so the websites and communication channels can be taken down. It argues that the use of its products and logos hurts its public image and has also filed other lawsuits against text-message scammers while endorsing several bills targeting scammers.
AI companies are under pressure to police bad actors using their platforms amid a wave of other scams and harmful chatbot interactions affecting young users. It is part of a growing backlash against the technology.
Last month, the online intelligence firm DarkTower identified a group of West African scammers who were using a sophisticated AI-powered system to send fake invoices to corporations, a type of fraud known as business email compromise.
The scammers built a tool that could manage their fraud operations from top to bottom, according to Gary Warner, director of threat intelligence at DarkTower. It conducts the initial research about the organization being targeted, discovers email accounts belonging to executives, then builds a phishing attack to steal their email login credentials, he said.
Once an email account is compromised, the AI system analyzes all of the messages and identifies ones related to payments or to people who have the authority to make payments in the organizations, Warner said. The AI system also recommends who to send them to and it writes the fake invoice emails as well.
Write to Robert McMillan at robert.mcmillan@wsj.com
(END) Dow Jones Newswires
June 12, 2026 05:00 ET (09:00 GMT)
Copyright (c) 2026 Dow Jones & Company, Inc.
Comments