← Back to feed
2026-06-06 PhishingMalicious Tool TLP:CLEAR

LLMShare: ChatGPT and Claude Share Links Weaponized for Malware via Google Ads and Fake Outage Pages

Date: 2026-06-06
Tags: phishing, malicious-tool

Executive Summary

ChatGPT share links host fake outage pages on OpenAI's genuine domain to distribute malware via Google ads, with attackers building pixel-perfect fake ChatGPT outage notices using ChatGPT's rendering capability. Push Security confirmed active detections on May 29, 2026. The technique exploits a design feature—not a software vulnerability—in ChatGPT's sharing system, meaning no patch from OpenAI can close it without changing product design.

Campaign Summary

FieldDetail
Campaign / MalwareLLMShare
AttributionUnknown (coordinated malvertising group) (confidence: low)
TargetChatGPT and Claude users searching for installation guides; geographically scoped and role-targeted via Google ads
VectorMalicious Google ads → chatgpt.com/s/ legitimate share link → fake outage page with custom HTML/CSS → download redirect to openew[.]app malware clone
Statusactive
First Observed2026-05-29

Detailed Findings

ChatGPT allows users to share conversations through public links in the format chatgpt.com/s/[unique-id], renders HTML and CSS code for developers to preview web layouts, and attackers used this rendering capability to build a pixel-perfect fake ChatGPT outage notice with OpenAI branding, polished error message, and prominent download button published as a standard share link. A parallel February 2026 campaign abused Claude.ai Artifacts to host fake Homebrew installation guides, hitting at least one confirmed government organization. LLMShare is part of a pattern that has become a defining characteristic of the 2026 threat landscape, with both ChatGPT and Claude variants surfacing in the same wave of customer detections suggesting a coordinated campaign or shared operational playbook actively experimenting across platforms. The download site at openew[.]app uses cloaking techniques to show malicious content only to specific targets; when security tools like URLScan visit, they see a harmless AR/VR website instead of the fake download page; the site offers downloads for both macOS and Windows that install malware.

MITRE ATT&CK Mapping

TechniqueIDContext
Spearphishing LinkT1566.002Malicious Google ads directing to trusted domain share links containing malware delivery lures
Fake CredentialsT1589.001Fake outage notice and cloned download portal impersonate legitimate OpenAI/Anthropic interfaces
MalwareT1588.001Trojanized desktop application delivered via cloaked download portal

IOCs

Domains

openew[.]app

Full URL Paths

chatgpt.com/s/ (legitimate domain, malicious share content)

Splunk Format

"openew[.]app" OR "chatgpt.com/s/ (legitimate domain, malicious share content)"

Detection Recommendations

Block malvertising by monitoring for ads to chatgpt.com/s/ and claude.ai/artifacts share links; alert on downloads of executables from chatgpt.com/s/ links (legitimate users do not download executables from share links); deploy email and browser warnings for downloads impersonating OpenAI/Anthropic; OSINT monitor for openew[.] and related typosquatter domains serving desktop app clones; threat intel teams: track cloaking infrastructure patterns used in this campaign for reuse in other LLM-based fraud.

References