The role of artificial intelligence in modern security technologies

Abstract

Abstract. Artificial intelligence continues to reshape security landscapes around the world, enabling systems that are more adaptive, predictive, and efficient than ever before. In recent months, several breakthroughs have highlighted both the opportunities and fresh threats posed by AI in the security domain. One unexpected vulnerability discovered in ChatGPT’s Deep Research agent, known as “ShadowLeak,” allows attackers to siphon sensitive information without any visible user interaction. This zero-click, server-side flaw underscores the fact that AI tools themselves can be weak links in enterprises’ security posture. At the same time, generative AI attacks are accelerating at an alarming rate. A Gartner report from September 2025 found that 29% of organizations experienced attacks targeting their AI application infrastructure within the past year. Deepfake incidents—through audio, voice biometric spoofing, or video—are growing in prevalence, introducing new vectors for disinformation and fraud. Another concern is that malware is now being developed that can evade even advanced protection tools. In one experiment, a malware strain trained via reinforcement learning managed to bypass Microsoft Defender’s security checks around 8% of the time after only three months of training, using a relatively small budget. This kind of AI-generated malware shows attackers’ growing ability to automate and refine their methods very quickly.