ESET's mid-2026 threat report paints a clear picture: threat actors are not abandoning proven attack techniques — they are adapting them to new platforms and user habits with remarkable speed. The report, released July 31, 2026, highlights four converging trends that are reshaping the malware landscape.
Malicious AI Skills Take Center Stage
The most notable emerging threat in the report is the weaponization of AI assistant plugins and skills. As enterprises and consumers increasingly rely on AI platforms — from Microsoft Copilot integrations to standalone AI assistants — attackers have identified the skill/plugin ecosystem as a new infection vector.
Malicious AI skills mimic legitimate productivity tools (file organizers, email drafters, code generators) while secretly:
- Exfiltrating sensitive data entered into AI prompts
- Harvesting credentials from connected applications
- Using AI platform permissions to access integrated cloud services
- Persisting through the AI platform's update mechanism
"The abuse of AI platforms represents a natural evolution," ESET researchers noted. "Attackers go where users are spending time and where trust is high."
AI-Assisted Malware Development Accelerates
Beyond targeting AI platforms, adversaries are using AI tools to accelerate their own development pipelines. ESET observed a measurable increase in malware samples showing characteristics of AI-generated code: consistent formatting, well-commented logic, and rapid variant production that suggests automated generation.
The practical impact: the barrier to entry for capable malware development continues to fall, and the volume of unique samples requiring analyst review continues to climb.
ClickFix Attacks Reach New Audiences
The ClickFix social engineering technique — which tricks victims into executing malicious commands by presenting fake browser or application errors — remains extremely active. ESET's telemetry shows ClickFix campaigns broadening their target base beyond IT-savvy users:
- Fake CAPTCHA pages presenting PowerShell execution instructions
- Bogus document rendering errors on cloud-hosted files (OneDrive, Google Docs)
- Impersonated IT support portals instructing users to "fix" issues by running scripts
- QR code lures on physical media left in public spaces, pointing to ClickFix pages
The technique's persistence reflects a fundamental truth: social engineering that instructs users to act — rather than exploiting software — bypasses most technical controls.
Record Quishing Activity
Quishing (QR code phishing) hit record levels in the first half of 2026. Attackers favor QR codes because:
| Advantage | Detail |
|---|---|
| Email filter bypass | QR codes are images — no embedded URLs for scanners to inspect |
| Mobile targeting | Scanning shifts the attack to personal devices with fewer controls |
| Fake legitimacy | QR codes on physical objects (parking meters, restaurant menus, conference badges) appear credible |
| MFA bypass potential | QR-based "re-authentication" lures capture real-time credentials |
ESET observed quishing campaigns impersonating Microsoft, DocuSign, and parcel delivery services at scale.
Ransomware Tools Engineered to Disable Security Software
The report documents a troubling refinement in ransomware tooling: purpose-built components designed to identify and terminate security software before encryption begins. These tools — often deployed via bring-your-own-vulnerable-driver (BYOVD) techniques — target EDR and antivirus solutions at the kernel level.
Notable observed capabilities include:
- Driver-based EDR termination using signed-but-vulnerable kernel drivers
- Process injection to hollow legitimate system processes and hide ransomware activity
- Shadow copy deletion executed via COM object hijacking to evade detection by VSS-monitoring tools
- Selective encryption targeting high-value file types while avoiding OS files to maintain system function
Recommendations
Organizations should address each of these trends directly:
AI skill hygiene: Audit installed AI plugins and skills across your environment. Establish an allowlist policy for AI integrations and review permissions granted to each skill.
ClickFix defense: User awareness training remains the primary control. Reinforce that IT support will never instruct users to run PowerShell commands from browser pop-ups. Disable PowerShell for standard users where possible.
Quishing protection: Deploy mobile threat defense (MTD) solutions that inspect URLs accessed from QR scans on managed devices. Train users to scrutinize QR code destinations before entering credentials.
EDR resilience: Ensure your EDR platform uses tamper protection and kernel-level self-defense. Monitor for BYOVD activity — legitimate driver loads at unusual times from non-standard paths.
Ransomware preparation: Maintain offline, immutable backups tested for recovery. Segment networks to limit ransomware spread and validate that shadow copy protection is monitored.