Google's Threat Intelligence Group has identified five AI-powered malware families that dynamically generate malicious scripts, evading detection. These malware programs utilize large language models like Gemini and Qwen to adapt in real-time, marking a shift from traditional malware development. Notably, families such as PROMPTFLUX and PROMPTSTEAL, associated with Russian and North Korean hackers, are using AI to rewrite code and create phishing content aimed at cryptocurrency exchanges. This development raises significant concerns about AI's role in cybercrime, with experts warning of increased threats to the crypto sector.