ESET's H1 2026 Threat Report indicates a significant increase in the use of artificial intelligence by attackers, both in developing malicious AI components and integrating AI into malware itself. The cybersecurity firm analyzed nearly 900,000 AI "skills"—functional components for AI agents—during the first half of 2026, identifying tens of thousands as suspicious and thousands as overtly malicious. This proliferation of AI skills is rapidly expanding the potential attack surface.
The report highlights PromptSpy, identified by ESET researchers as the first known Android malware to incorporate generative AI into its execution flow. PromptSpy leverages Google's Gemini large language model to interpret user interface elements, allowing it to adapt across various devices and environments without relying on hardcoded behaviors. While still uncommon, this development suggests a future of more flexible and adaptable threats, though built-in safeguards in LLMs may be slowing broader adoption.
Beyond AI, ESET observed attackers refining established techniques for greater efficiency and scalability. The "ClickFix" social engineering technique, which uses fake error messages, has expanded beyond fake CAPTCHA prompts. It now appears in AI-themed help pages, browser extensions, and cloud authentication scenarios. ESET's detections of ClickFix more than doubled between the second half of 2025 and the first half of 2026, indicating sustained and evolving activity.
Phishing campaigns are also adapting to user behavior, with QR code phishing, or "quishing," reaching record levels in ESET's telemetry. Attackers are embedding malicious links in QR codes to bypass initial security checks and direct users to mobile devices, exploiting the inherent trust many place in these codes.
Ransomware activity remains robust, with attackers continuing to employ "EDR killers"—tools designed to disable security software during attacks. ESET Research has documented over 100 distinct EDR killer variants in use, with new ones emerging regularly. However, data from various sources suggests a declining trend in victims paying ransoms, which may indicate improved mitigation and response strategies.
