ESET Threat Report H1 2026: AI Skills Explode, PromptSpy Android Malware Emerges
ESET's H1 2026 threat report highlights the rapid growth of AI skills, the emergence of AI-powered Android malware like PromptSpy, and evolving phishing tactics including record quishing levels.
The cybersecurity landscape in the first half of 2026 has seen threat actors significantly enhance the efficiency and scalability of their operations, not by inventing entirely new methods, but by adeptly adapting established techniques to new platforms, technologies, and user behaviors. This evolution is increasingly driven by the growing role of artificial intelligence, which is rapidly expanding the attack surface.
ESET's analysis revealed a surge in AI skills – small functional components used by AI agents – with nearly 900,000 instances identified. Tens of thousands of these were deemed suspicious, and thousands outright malicious. The sheer volume and rapid growth of these AI skills within the burgeoning AI ecosystem present a significant new frontier for potential exploitation.
AI is also beginning to infiltrate malware itself. Following the advent of AI-powered ransomware in 2025, ESET researchers have identified PromptSpy, the first known Android malware to incorporate generative AI into its execution flow. PromptSpy utilizes Google's Gemini to interpret user interface elements, allowing it to adapt dynamically across different devices and environments without relying on hardcoded behaviors. While its current prevalence is low, PromptSpy serves as a stark indicator of the potential for more flexible and evasive future threats, though built-in safeguards in large language models may temper this adoption.
The social engineering technique known as ClickFix, which employs fake error messages, has expanded its reach beyond simple CAPTCHA prompts. Attackers are now leveraging this method within AI-themed help pages, browser extensions, and cloud authentication scenarios. ESET observed a more than doubling of detections for this vector between the latter half of 2025 and the first half of 2026, underscoring its sustained activity and adaptability.
Phishing campaigns are also evolving in lockstep with user behavior. QR code phishing, or quishing, has reached unprecedented levels according to ESET's telemetry. Attackers are embedding malicious links within QR codes to bypass initial scrutiny and redirect user interaction to mobile devices, exploiting the implicit trust many users place in these seemingly innocuous black-and-white squares.
Ransomware activity shows no signs of abating, with attackers continuing to deploy EDR (Endpoint Detection and Response) killers to disable security software during attacks. ESET Research has documented over 100 distinct EDR killers in the wild, with new variants emerging regularly. Despite this, data indicates a declining proportion of victims are choosing to pay ransoms, suggesting some progress in global mitigation and response strategies.
The report also touches on the broader implications of AI in cybersecurity, noting how AI coding agents can inadvertently trigger endpoint security rules by mimicking attacker behavior, and how vulnerabilities like 'GhostApproval' can allow AI coding assistants to bypass sandboxes. Furthermore, the emergence of botnets marketed as 'AI-as-a-Service' like the Mycelium Framework signals a new era of AI-powered cybercrime infrastructure.
Overall, ESET's H1 2026 report paints a picture of a rapidly evolving threat landscape where AI is not just a tool for defense but a potent enabler for sophisticated and adaptive cyberattacks, demanding continuous vigilance and innovation in security measures.