Google Sues a Scam Network That Built Phishing Sites With Gemini

Google has gone to court against the people it says turned its own AI against the public. On June 12 the company filed suit against an alleged China-based criminal operation it calls the "Outsider Enterprise," accusing it of running a phishing-as-a-service platform that let scammers spin up convincing fake websites at industrial scale. According to the complaint, the group leaned on Gemini and other AI tools to write the website code, then imported it into its own software to churn out the finished traps.

The targets were familiar by design. As Engadget reported, the fake sites impersonated Google, YouTube, and government services including the US Postal Service and New York's E-ZPass toll system. These are the same brands behind the "you have an unpaid toll" and "your package is held" texts that have flooded phones over the past year.

The numbers are the story

What stands out is the scale that automation made possible. Google says it detected more than 1.59 million URLs tied to the operation between November 2025 and April 2026, spanning over 9,000 distinct fake websites. During a two-week stretch in May, Android users flagged roughly 55,000 spam texts linked to the campaign. The company estimates hundreds of thousands of victims and losses in the millions.

Those figures are the point. A human team can build a few dozen passable phishing pages. A team with a model writing the code can build thousands, refresh them as fast as they are blocked, and tune each one to a specific brand. As Bloomberg noted, the AI did not invent the scam, but it removed the friction that used to cap how big a scam could get.

A first for Google

Google said this is the first time it has taken legal action against bad actors for misusing Gemini specifically to defraud US consumers. The case was filed alongside cooperation with the FBI and major telecom carriers, part of a broader crackdown on AI-assisted fraud.

It also fits a pattern. Only days ago OpenAI disrupted China-linked accounts using ChatGPT for influence operations. The labs are increasingly forced to police the misuse of their own products, and litigation is the newest tool on the shelf. Whether a lawsuit against an operation hidden behind shell layers in another country actually stops the fraud is the open question. What it does signal is that "someone misused our model" is no longer a defense the companies are willing to leave unanswered.