A Chrome extension that Google pulled from its store in January 2026 over conversation-theft allegations is back in circulation and reaching enterprise browsers again through Google’s own CRX distribution infrastructure, according to Netskope Threat Labs.
The extension is named ‘AI Sidebar with DeepSeek, ChatGPT, Claude and more’ and is advertised as an AI assistant for Chrome. OX Security flagged it in December 2025 for scraping ChatGPT and DeepSeek conversation content and sending it to external domains. Prior to its removal, the extension amassed over 300,000 installs and a 4.6-star rating on the Chrome Web Store.
The malicious behavior disappeared from later builds, and the extension’s privacy policy now references the earlier conduct. But the reprieve was temporary. As of August 2026, the extension is active on the Chrome Web Store again, and Netskope says enterprise endpoints are currently receiving updates carrying a different kind of payload.
Version 1.7.2.0, distributed between July 20 and 31, 2026, shipped clean functionality, seemingly designed to establish a trustworthy update history. Version 1.7.3.0 followed roughly two weeks later with what Netskope describes as a surgical 21-line addition to a single script file.
The uninstall path exploits a Chrome quirk where only the most recently registered uninstall URL fires. The operators overwrite the legitimate code’s URL five seconds after it loads, ensuring that users who remove the extension still generate a referral on the way out.
Attribution points to a company called Extchange.com, listed in the extension’s own files as both developer and data controller, with a domain registered in February 2024 and no public registrant information. The Chrome Web Store listing falsely shows the developer name as DeepSeek AI.
Netskope says it detected and blocked version 1.7.3.0 arriving at enterprise endpoints via Google’s CDN, classifying it as Trojan.GenericFCA.Script.37952. The company reported the extension to the Chrome Web Store before publishing its findings and is advising organizations to remove it.
Netskope frames the affiliate scheme itself as low-stakes compared with what made the case notable. However, the security firm warned that the same channel could just as easily carry a more damaging payload next time.
Related: Critical Flaws Discovered in Belgian eID Software Used by 2 Million People
Related: ‘Ghostjacking’ Attack Uses Poisoned Logs to Turn AI Agents Bad
Related: CISA Urges Immediate Patching of Exploited Progress LoadMaster Vulnerability