Threat Actors Abuse ChatGPT Share Links to Host Fake Outage Pages Delivering Malware

Threat Actors Abuse ChatGPT Share Links to Host Fake Outage Pages Delivering Malware

What Happened — Threat actors are leveraging OpenAI’s “ChatGPT Share” feature to publish malicious pages that mimic an official outage notice. When users click the “Download desktop app” button they are redirected to a counterfeit OpenAI download site that serves Windows and macOS malware.

Why It Matters for TPRM —

• The abuse occurs on a legitimate chatgpt.com domain, bypassing many URL‑based defenses.
• Malware distribution via a trusted SaaS platform expands the attack surface of any organization that permits employee use of AI tools.
• Similar tactics have been observed on other AI platforms, indicating a growing supply‑chain style threat vector.

Who Is Affected — SaaS/AI providers, enterprises that integrate ChatGPT into workflows, and end‑users across all industries.

Recommended Actions —

• Block or closely monitor outbound traffic to chatgpt.com/s/* URLs.
• Educate users that OpenAI does not distribute desktop installers via shared links.
• Deploy web‑gateway URL‑reputation controls that flag redirects to unknown domains (e.g., openew.app).
• Review third‑party risk assessments for AI SaaS vendors to ensure they cover content‑sharing abuse scenarios.

Technical Notes — The campaign (named “LLMShare”) uses Google Ads to drive search traffic to a crafted shared link (chatgpt.com/s/...). The page renders custom HTML/CSS via ChatGPT’s rendering engine, displaying a fake outage banner and a download button. The download redirects to openew.app, which hosts cloaked malware binaries that perform VM detection and install infostealers. Similar abuse has been reported on Claude (Anthropic) and Grok (Microsoft). Source: BleepingComputer