# FBI/Google Disrupt 'Outsider Enterprise' AI-Powered (Gemini) Phishing-as-a-Service and Smishing Network (Operation Riptide / Operation Ghost Hook)

> Outsider Enterprise was a China-based phishing-as-a-service (PhaaS) and mass-smishing syndicate, active since July 2023, that sold a Telegram-distributed phishing kit for as little as $88/week and coached customers to weaponize Google Gemini and other AI tools to generate convincing fake brand pages. It is linked to ~9,000 fake sites, over 1 million fraudulent URLs, ~3.87 million stolen credit cards and an estimated $1.9 billion in losses before a June 12, 2026 takedown by the FBI, Google, and Lumen.

- **Published:** 2026-06-14T00:00:00Z
- **Last reviewed:** 2026-06-14T00:00:00Z
- **Canonical:** https://intel.threadlinqs.com/threat/TL-2026-0787
- **ID:** TL-2026-0787
- **Severity:** HIGH
- **Category:** PHISHING
- **Status:** RESOLVED
- **Actor:** Outsider Enterprise (China)
- **Detections:** 9 · **IOCs:** 16 (full data via the Threadlinqs MCP server — Purple tier)

## Description

Outsider Enterprise operated one of the largest commodity phishing-as-a-service (PhaaS) and SMS-phishing (smishing) ecosystems documented in the United States. Coordinated entirely through Telegram and based in China, the operation lowered the technical bar for fraud: subscribers paid as little as $88 per week (or $200 per month) for a self-service kit that included 290+ pre-built fraudulent website templates impersonating trusted brands, a campaign-performance dashboard, and built-in keystroke logging to capture victim input in real time. License sales and provisioning were automated through a Telegram bot, @OutsiderCodeBot.

A defining feature of the operation was the deliberate weaponization of generative AI. According to Google's civil complaint, Outsider encouraged and provided step-by-step instructions for customers to prompt Google Gemini (and other AI platforms) to write the HTML for phishing pages — for example, framing the request as building a benign 'gift redemption page,' instructing the model to avoid JavaScript and to use inline CSS, then copying the generated code into Outsider's interface. This prompt-laundering approach used innocuous framing to evade AI safety guardrails and rapidly mass-produce convincing brand-impersonation pages.

The enterprise was structured into five interconnected divisions: a Developer Group (software and templates), a Data Broker Group (targeting/victim lists), a Spammer Group (bulk SMS tooling), a Theft Group (monetizing and laundering stolen data), and a Telegram Group (coordination and recruitment). Smishing lures impersonated USPS/missed-package notices, unpaid toll and highway-violation notices (e.g., E-ZPass), parking violations, brokerage-account problems, wireless-carrier rewards, and Google itself, steering recipients to credential- and payment-card-harvesting sites. Over a two-week window (May 18-June 1, 2026), roughly 2.5 million scam texts were sent to Android users, 55,000 of which were flagged as spam.

The disruption — branded Operation Ghost Hook within the FBI's broader Operation Riptide — was announced June 12, 2026. The FBI, in partnership with Google, Lumen Technologies (Black Lotus Labs), and carriers AT&T, T-Mobile, and Verizon, seized multiple administration servers, a Shopify e-commerce storefront, a threat-actor testing account, the Telegram bot containing customer data, and approximately $100,000 in USDT from payment wallets, and rerouted thousands of U.S.-registered phishing domains to an FBI splash page. Google simultaneously filed a civil lawsuit in Manhattan federal court. The operation parallels the separately disrupted China-based 'Lighthouse' PhaaS platform (November 2025). No software CVE is involved; this is a criminal-service and abuse-of-AI threat, scored HIGH on the basis of its scale and confirmed financial impact.

## MITRE ATT&CK

- T1583 Acquire Infrastructure
- T1587 Develop Capabilities
- T1608 Stage Capabilities
- T1588 Obtain Capabilities
- T1598 Phishing for Information
- T1589 Gather Victim Identity Information
- T1566 Phishing
- T1660 Phishing
- T1056 Input Capture
- T1056 Input Capture
- T1684.001 Impersonation
- T1036 Masquerading
- T1102 Web Service
- T1071 Application Layer Protocol
- T1567 Exfiltration Over Web Service
- T1657 Financial Theft

## Sources

- [FBI disrupts massive AI-powered phishing service using a million URLs](https://www.bleepingcomputer.com/news/security/fbi-disrupts-massive-ai-powered-phishing-service-using-a-million-urls/)
- [Google Sues Chinese Smishing Network Accused of Using Gemini AI in Phishing](https://thehackernews.com/2026/06/google-sues-chinese-smishing-network.html)
- [FBI takes down massive China-based cybercrime network that caused $1.9B in losses](https://cyberscoop.com/outsider-cybercrime-network-takedown-china-fbi-google-lumen/)
- [How Google is combatting AI scams and dismantling the 'Outsider Enterprise'](https://blog.google/innovation-and-ai/technology/safety-security/combatting-ai-scams/)
- [Chinese cybercrime operation that used AI to scam 'hundreds of thousands of victims' sued by Google](https://techcrunch.com/2026/06/12/chinese-cybercrime-operation-that-used-ai-to-scam-hundreds-of-thousands-of-victims-sued-by-google/)
- [Google sues suspected AI scam network (Washington Examiner)](https://www.washingtonexaminer.com/policy/technology/4605495/google-ai-phishing-lawsuit-gemini-scams/)
- [Google Sues China-Based Cybercrime Ring That Used AI to Scam 100,000+ Americans (Cyber Kendra)](https://www.cyberkendra.com/2026/06/google-sues-china-based-cybercrime-ring.html)

## Full data

Detection queries (Splunk SPL / Microsoft KQL / Sigma) and IOC values require the Threadlinqs MCP server (Purple tier): https://intel.threadlinqs.com/mcp

Canonical: https://intel.threadlinqs.com/threat/TL-2026-0787