BBLabs NewsBBLabs News
NewsAll articlesTopics
ES
BBLabs NewsBBLabs News

BBLabs News

Una historia al día. Cero ruido.

Newsletter técnica de ciberseguridad. Una historia al día sobre CVEs críticos, brechas, bug bounty e IA. Filtrado por IA, escrito para humanos.

Producto

  • Hemeroteca
  • Ediciones
  • Temas
  • Glosario
  • RSS
  • Atom
  • JSON Feed

Editorial

  • Acerca de
  • Suscribirse
  • Cuenta
  • English

Legal

  • Privacidad
  • Términos
  • Contacto: team@bblabs.es

Conectar

  • YouTube · @0xGorka
  • Instagram · @bblabs.es
  • Discord BBLabs
  • Discord Bug Bounty ES
21 artículos·5 ediciones·Desde 2026·Hecho en España
© 2026 BBLabs News·Por Gorka El Bochi
BBLabs NewsBBLabs News
NewsAll articlesTopics
ES
ChatGPT share links abused to deliver malware
Back to homeIA

ChatGPT share links abused to deliver malware

Threat actors are abusing ChatGPT share links to serve fake OpenAI outage pages that deliver malware disguised as the desktop app.

  1. Home
  2. ›
  3. IA
  4. ›
  5. ChatGPT share links abused to deliver malware
by Gorka El Bochi Morillo
·
2 min read
·May 31, 2026

What happened

Threat actors are *exploiting* ChatGPT's conversation-sharing feature to stage fake OpenAI outage pages. The attack is straightforward: craft a conversation whose content mimics an official service-down notice, share it via a `chat.openai.com/share/…` link, and distribute that link through email, social media, or forums.

The target lands on an OpenAI domain, sees a page with the correct branding and a "service unavailable" message, and gets offered a "ChatGPT desktop client" download to keep accessing the service. The installer is malware. The exact payload type isn't detailed in the original report, but the download vector is clear.

What makes this particularly effective: the link lives on `chat.openai.com`. Valid HTTPS, high-reputation domain. Corporate email filters, proxies, and domain-reputation-based security tools don't flag that hostname.

Why it matters

Most phishing campaigns rely on lookalike domains registered days before the attack — easy to catch by domain age or low reputation. Here the malicious content is hosted on OpenAI's own infrastructure. Reputation controls don't apply. A user who would immediately distrust `openai-support-help[.]xyz` won't hesitate on `chat.openai.com`.

Confusion around ChatGPT desktop availability amplifies the deception. Many users don't know exactly which platforms have a native client, so an "install this to keep using the service" prompt doesn't trigger alarm. The chain — *legitimate platform → fake outage notice → trojanized installer* — is a well-established pattern seen across GitHub, Google Docs, and Notion, but brand trust around AI tools drives the hit rate higher.

What to do

  • Only download ChatGPT Desktop from `openai.com/chatgpt/download`
  • Treat any `chat.openai.com/share/…` link that prompts a software install as malicious by default
  • Configure your proxy or EDR (endpoint detection and response tool) to alert on executables downloaded from AI-platform domains
  • If you manage non-technical users, communicate explicitly: ChatGPT share links are not an official software distribution channel

The root issue isn't an OpenAI vulnerability — it's a design abuse. Any platform with user-generated content that renders public pages becomes a potential payload host. Until OpenAI filters or restricts what's renderable in public shares, this attack surface stays open.

What to do

  • Download ChatGPT Desktop only from `openai.com/chatgpt/download`
  • Block executable downloads originating from AI-platform share links in your proxy
  • Brief non-technical users: ChatGPT share links never push official software

Share this story

Help more people discover BBLabs News.

ChatGPT share links abused to deliver malware
VerticalDownload image
LinkedInXWhatsApp

Interested in IA?

Subscribe to this stream and get the most relevant news every day — no spam, no noise.

Subscribe

Related articles

Destacado
IA30 may 2026·2 min

Anthropic's Mythos: 23K vulns found in 1,000 OSS projects

Anthropic's Mythos AI agent detected 23,000 potential vulnerabilities across 1,000 open source projects, many already confirmed critical.

  • Run `trivy` or `syft` on your full dependency tree before the next sprint.
  • Subscribe to OSV.dev for automated CVE alerts across your OSS dependencies.
  • Enable SBOM generation in your build pipeline to react fast when Mythos CVEs drop.
Gorka El Bochi Morillo
Leer artículo
IA26 may 2026·2 min

RAMPART & Clarity: security testing for AI agents

Microsoft open-sources RAMPART and Clarity, two frameworks for security-testing AI agents at development time.

Leer artículo
IA26 may 2026·2 min

Anthropic's restricted Mythos model may ship inside Claude Code

Anthropic is preparing to roll out Mythos to Claude Code, a restricted model flagged for major security risks to public and private software.

Leer artículo

Want to get news like this every day?

Browse all articles
BBLabs NewsBBLabs News

BBLabs News

Una historia al día. Cero ruido.

Newsletter técnica de ciberseguridad. Una historia al día sobre CVEs críticos, brechas, bug bounty e IA. Filtrado por IA, escrito para humanos.

Producto

  • Hemeroteca
  • Ediciones
  • Temas
  • Glosario
  • RSS
  • Atom
  • JSON Feed

Editorial

  • Acerca de
  • Suscribirse
  • Cuenta
  • English

Legal

  • Privacidad
  • Términos
  • Contacto: team@bblabs.es

Conectar

  • YouTube · @0xGorka
  • Instagram · @bblabs.es
  • Discord BBLabs
  • Discord Bug Bounty ES
21 artículos·5 ediciones·Desde 2026·Hecho en España
© 2026 BBLabs News·Por Gorka El Bochi