On June 12, 2026, Google filed a complaint in the Southern District of New York against "Outsider Enterprise," a China-based cyber network that converted generative AI into an industrial component of a phishing-as-a-service platform. The case marks the first time a Big Tech company has pursued civil litigation to strike at the structured abuse of its LLM by organized criminals, with the FBI and major U.S. telecom carriers providing operational support.
- The Outsider Enterprise network used Gemini to generate HTML/CSS code for phishing sites, integrating it directly into its criminal-as-a-service platform
- The platform offers more than 290 templates impersonating Google, YouTube, USPS, financial institutions, and U.S. government agencies
- According to Google, 2.5 million smishing messages were sent to Android users in two weeks, generating over 55,000 spam text reports
- The FBI estimates 3.87 million credit cards stolen and $1.9 billion in losses since July 2023, with roughly 100,000 USDT seized in the parallel "Ghost Hook" operation
How Gemini Became the Engine of a Phishing Factory
According to Google's dossier, Outsider Enterprise members "actively encouraged each other" to use Gemini to generate custom phishing page code. The technique relies on prompt engineering: requests framed as harmless assistance creating "royal redemption pages" that actually produce inline HTML/CSS artifacts ready for integration into the Outsider suite.
The network is structured into five interconnected groups with specialized functions: Developer, Data Broker, Spammer, Theft, and a Telegram coordination group. The phishing kit costs $88 per week or $200 per month, purchasable via the Telegram bot @OutsiderCodeBot. The dashboard includes performance tracking and real-time keystroke logging.
Google has disabled the Gemini accounts and infrastructure linked to the abuse. The company does not specify, however, which technical controls have been strengthened to prevent similar future exploits beyond the revocation of identified accounts.
The Numbers of an Industrial Campaign: From 9,000 Sites to Millions of URLs
The scale of the Outsider Enterprise operation exceeds the parameters of traditional phishing campaigns. According to data reported by Google, the network generated more than 9,000 fake sites and over one million fraudulent URLs linked to the service. In a two-week span — from May 18 to June 1, 2026 — smishing messages reached 2.5 million sends to Android users, with over 55,000 spam texts reported and an average exceeding two complaints per minute.
"The operation weaponized Gemini to help generate fraudulent phishing pages and deploy massive SMS phishing ('smishing') attacks"
— Google, cited by The Hacker News
Victim estimates range from hundreds of thousands according to Google to far higher aggregate figures in the FBI's extended timeline. The Bureau estimates 3.87 million credit cards stolen and $1.9 billion in losses since July 2023. In the parallel "Ghost Hook" operation, the FBI seized domains and roughly 100,000 USDT from wallets linked to the network.
An Unprecedented Legal Front: RICO Act and AI Platform Liability
Google's complaint invokes two U.S. federal statutes: the RICO Act, traditionally used against structured criminal organizations, and the Lanham Act, which protects trademarks from fraudulent use. The choice of RICO is particularly significant: it applies a criminal framework for civil purposes against a network operating overseas, sidestepping jurisdictional difficulties in pursuing unidentified individuals.
The filing's sign-off by General Counsel DeLaine, cited by the New York Times via Cyber Security News, confirms the strategic nature of the action. Google is not merely seeking damages; it aims to establish that abuse of generative AI platforms constitutes grounds for executable civil actions, with the platform itself as an injured party rather than a bystander.
The case sits in a regulatory void that remains open. No specific federal law currently governs LLM provider liability for outputs subsequently used in criminal activity. Google backs seven bipartisan proposals, including the Stop SCAMS Act, which would steer Congress toward a national anti-fraud framework.
Why It Matters
The dossier does not document specific corrective measures adopted by Gemini to prevent future abuse beyond the disabling of identified accounts. The source also does not specify whether other AI platforms beyond Gemini were used by the network, nor does it identify the personal names of the operators behind Outsider Enterprise.
The judicial outcome of the complaint is pending at the time of reporting, with no certain timeline for a decision. The exact number of victims and total verified losses remain undetermined: available figures are estimates, not audited totals. It also does not emerge whether Google suffered direct financial damages beyond brand reputation erosion.
The technical angle of the case — LLMs as a generative module inside criminal-as-a-service platforms — raises questions the brief does not address textually: the nature of data exposed by victims, the possible presence of mechanisms evading Gemini's security controls, and the frequency with which engineered prompts bypass moderation filters.
Carrier Coordination and the FBI's Role
The response to Outsider Enterprise involved AT&T, T-Mobile, and Verizon in a carrier-level message blocking action — a rare cooperation among direct competitors under a tech company's coordination. Brett Leatherman, Assistant Director of the FBI Cyber Division, stated that "criminals use AI to make fraud more convincing and harder to detect." According to the same source, the perpetrators "built a business impersonating trusted brands to defraud hundreds of thousands of victims."
Google's defensive model integrates AI detection against AI-generated fraud: an operational paradox that increasingly characterizes 2026 cybersecurity. The official blog reports that Google defenses intercept over 10 billion malicious messages per month, an order of magnitude that justifies investment in automated tools but does not eliminate the risk of false negatives on prompt-engineered campaigns.
FAQ
Why did Google sue instead of limiting itself to a technical response?
The civil route serves to establish a legal precedent on liability for generative AI platform abuse, beyond obtaining injunctions and seizures that a technical response alone cannot achieve.
Did China officially sponsor the operation?
The brief reports the network is "based in China," but no evidence of government sponsorship emerges. "Based in China" does not equate to a nation-state operation.
Does Gemini have a security vulnerability?
No. The abuse occurs via legitimate accounts with engineered prompts designed to appear harmless. It is not a technical vulnerability of the model, but a functional exploit of its generative capabilities.
Sources
- https://cybersecuritynews.com/google-sues-chinese-cybercrime-network/
- https://thehackernews.com/2026/06/google-sues-chinese-smishing-network.html
- https://blog.google/innovation-and-ai/technology/safety-security/combatting-ai-scams/
- https://thehackernews.com/
- https://thehackernews.com/p/upcoming-hacker-news-webinars.html
- https://thehackernews.com/search/label/Threat%20Intelligence
- https://thehackernews.com/search/label/Vulnerability
- https://thehackernews.com/search/label/Cyber%20Attack
Information is based on cited sources and current as of publication.