Cybersecurity researchers have reported the discovery of two new malicious extensions on the Chrome Web Store. These extensions were specifically designed to stealthily exfiltrate conversation histories from major generative AI services, including OpenAI ChatGPT and DeepSeek, alongside the users’ general browsing data, transmitting them to servers controlled by the attackers.
A combined total of over 900,000 users have been affected, putting them at risk of large-scale data leakage. One of the identified extensions, titled “Chat GPT for Chrome with GPT-5, Claude Sonnet & DeepSeek AI,” leveraged the promise of accessing popular AI models to lure a large user base.
As the use of generative AI becomes ubiquitous, incidents involving the theft of sensitive data via browser extensions that users unknowingly install are on the rise. Since extensions often possess access permissions to nearly all data accessed by the browser, malicious code can easily steal highly confidential interactions with AI (such as corporate information, personal data, or project details).
From a Web3 security perspective, this is considered a critical incident, as conversations with AI models may contain sensitive information pertaining to smart contract management or private key handling. Users are strongly urged to rigorously scrutinize the provider’s credibility, reviews, and requested permissions before installing any browser extension.
ソース元: Two Chrome Extensions Caught Stealing ChatGPT and DeepSeek Chats from 900,000 Users
コメント