Escalating AI-Driven Cyber Threats: Browser Ransomware, Advanced Prompt Injection, and Supply Chain Risks Emerge

Key Intelligence

• ■New "browser-only" ransomware attacks are being generated via LLM hallucinations, exploiting Chromium APIs to lock user browsers on Windows and Android.
• ■Advanced prompt injection techniques like "CoT Forgery" and "InkJect" (visual prompt injection) are effectively bypassing LLM guardrails to extract sensitive or forbidden information.
• ■Searches for AI "prompt injection" attacks have more than doubled in the past year, signaling a significant increase in this attack vector.
• ■AI is facilitating novel supply chain threats such as "Phantom Squatting" and introduces new risks through LLM aggregators and AI API proxies.
• ■These developments underscore critical security vulnerabilities in AI-powered browsers and the broader deployment of Large Language Models, demanding urgent mitigation strategies.