AI agents stole 190k gaming credentials for under $400 in model costs (research.checkpoint.com)

Recent research by Check Point revealed a concerning trend in cybercrime where AI-driven hacking platforms like ARTEX are being exploited by criminals to steal gaming credentials and other sensitive data for minimal costs—around $400 in model expenses. ARTEX, initially designed for ethical penetration testing, has over 1,800 publicly accessible instances, many operated almost autonomously after the initial goal setup. This disturbing shift towards autonomous cyber attacks allows threat actors to deploy significant attacks with little technical knowledge, making it easier for a broader audience to engage in opportunistic cybercrime. The research documented several campaigns utilizing ARTEX, with notable activities including the theft of Apple’s iOS code-signing certificates, where agents operated with minimal human intervention. The tools exploited various web vulnerabilities like Remote Code Execution (RCE) to accomplish their illicit goals. The prevalence of Chinese AI models in these attacks raises alarms about the potential for widespread, scalable exploitation in the AI/ML space. This situation underscores the urgent need for enhanced cybersecurity measures as autonomous AI systems evolve from beneficial tools into platforms for sophisticated cybercriminal endeavors.