Malicious Browser Extensions Hijack AI Chat Data in "Prompt Poaching" Campaigns Security researchers have uncovered a surge in malicious browser extensions designed to steal sensitive AI chat data a tactic now called "prompt poaching." These extensions exploit the growing use of AI-powered tools in browsers, intercepting user prompts and responses in real time without detection. ### How the Attack Works AI browser extensions, marketed for seamless integration with platforms like chatbots, often request broad permissions to read page content and monitor activity across tabs. While legitimate versions enhance productivity, compromised or fake extensions abuse these permissions to scrape AI conversations via DOM manipulation or API interception. The stolen data including business intelligence, credentials, or proprietary code is then sent to attacker-controlled servers. ### Distribution Methods Threat actors deploy two primary tactics: 1. Cloned Extensions – Malicious versions of popular tools (e.g., those mimicking AITOPIA’s extensions) are distributed, often indistinguishable from the originals. 2. Supply Chain Attacks – Legitimate extensions with large user bases (e.g., Urban VPN Proxy) are hijacked post-installation, introducing hidden data-collection features. ### Impact and Risks Prompt poaching poses severe risks for both individuals and enterprises: - Data Exposure: Stolen AI conversations may contain confidential business data, internal communications, or customer information. - Phishing & Monetization: Attackers can weaponize captured prompts for targeted phishing or sell them on underground markets. - Enterprise Threats: Employees using unvetted extensions may inadvertently leak intellectual property or sensitive workflows. ### Mitigation Efforts Security experts recommend: - Restricting unapproved extensions via browser management tools or group policies. - Prioritizing official AI tools from trusted vendors over third-party extensions. - Auditing permissions and monitoring for unusual outbound connections to detect malicious activity. As AI adoption grows, prompt poaching underscores the need for heightened scrutiny of browser-based tools, where convenience often comes at the cost of security.