By 
Cybersecurity researchers have uncovered two malicious Chrome extensions that secretly siphon chatbot conversations and browsing data to servers controlled by attackers.
The extensions — together installed by more than 900,000 users — are:
- Chat GPT for Chrome with GPT-5, Claude Sonnet & DeepSeek AI (600,000 users)
- AI Sidebar with Deepseek, ChatGPT, Claude, and more. (300,000 users)
According to researchers at OX Security, the extensions masquerade as productivity tools while quietly sending ChatGPT and DeepSeek chats, along with the URLs of all open tabs, to remote command-and-control servers every 30 minutes. Users are misled into approving “anonymous analytics,” but the tools actually extract full conversation content.
The extensions imitate a legitimate tool called “Chat with all AI models (Gemini, Claude, DeepSeek…) & AI Agents” from AITOPIA, which has about 1 million users. Both malicious copies remain available in the Chrome Web Store, although one has lost its “Featured” badge.
Once installed, the malware requests permission to collect anonymized behavior data. If granted, it scrapes webpage elements to capture chatbot messages, stores them locally, and later uploads them to domains like chatsaigpt[.]com and deepaichats[.]com. The attackers even used the AI web-development platform Lovable to host privacy policies and infrastructure, further disguising their intent.
The risk is significant: stolen chats might include corporate data, personal details, internal URLs, and search activity — all of which could be exploited for espionage, identity theft, phishing, or resale on underground markets.
“Prompt Poaching” Expands — Even to Legitimate Tools
The discovery comes as Secure Annex reports that some popular, legitimate extensions — including Similarweb and Sensor Tower’s Stayfocusd — also collect AI prompts and outputs as part of their analytics. Updated terms and privacy notices now explicitly state that data entered into AI tools may be gathered and analyzed.
Researchers found these tools may scrape page DOM elements or intercept browser APIs to capture prompts from platforms such as ChatGPT, Claude, Gemini, and Perplexity. Secure Annex warns this practice, dubbed “Prompt Poaching,” is becoming a lucrative trend — and developers may increasingly adopt it as a monetization strategy.
The users who installed the suspicious extensions must immediately remove them from their browser. Avoid installing unknown extensions, even if labeled “Featured.” Always review extension permissions carefully — especially those requesting access to browsing activity or chatbot content.
As Secure Annex notes, this is likely only the beginning: browser extensions are fast becoming a prime vector for capturing sensitive AI conversations — and users need to stay vigilant.
