Particle.news

Researchers Find 16 Malicious ‘ChatGPT’ Extensions Stealing Session Tokens

The operation abuses high‑privilege browser access to skim authentication tokens that grant attackers entry to users’ ChatGPT sessions.

Overview

  • LayerX researchers identified at least 16 browser add‑ons posing as ChatGPT productivity tools, with 15 on the Chrome Web Store and one on the Microsoft Edge marketplace.
  • A script injected into chatgpt.com monitors outbound requests to exfiltrate session tokens along with metadata, usage telemetry, and backend‑issued access tokens.
  • Stolen tokens let attackers impersonate victims to access chat histories and connected services such as Slack and GitHub.
  • The extensions share code, branding, upload timing, and infrastructure linked to domains including chatgptmods.com and Imagents.top, indicating a single coordinated actor.
  • Roughly 900 installs have been recorded so far and the listings remain available, prompting researchers to urge users to remove suspicious AI‑related extensions as one even displayed a Chrome Web Store “featured” badge.