Adversaries weaponising and targeting AI at scale: report

CrowdStrike has released its 2025 Threat Hunting Report, highlighting a new phase in modern cyber attacks in which adversaries are operationalising GenAI to scale operations and accelerate attacks — and increasingly targeting the autonomous AI agents reshaping enterprise operations. The report reveals how threat actors are targeting tools used to build AI agents — gaining access, stealing credentials, and deploying malware — a clear sign that autonomous systems and machine identities have become a core part of the enterprise attack surface.

Report highlights

Based on frontline intelligence from CrowdStrike’s elite threat hunters and intelligence analysts tracking more than 265 named adversaries, the report reveals the following five issues.

Adversaries weaponising AI at scale

DPRK-based adversary FAMOUS CHOLLIMA used GenAI to automate every phase of its insider attack program. From building fake resumes and conducting deepfake interviews to completing technical tasks under false identities, AI-powered adversary tradecraft is transforming traditional insider threats into scalable, persistent operations.

Russia-based adversary EMBER BEAR used GenAI to amplify pro-Russia narratives and Iran-based adversary CHARMING KITTEN deployed LLM-crafted phishing lures targeting US and EU entities.

Agentic AI is the new attack surface

CrowdStrike observed multiple threat actors exploiting vulnerabilities in tools used to build AI agents, gaining unauthenticated access, establishing persistence, harvesting credentials, and deploying malware and ransomware. These attacks demonstrate how the agentic AI revolution is reshaping the enterprise attack surface — turning autonomous workflows and non-human identities into the next frontier of adversary exploitation.

GenAI-built malware becomes real

Lower-tier eCrime and hacktivist actors are abusing AI to generate scripts, solve technical problems, and build malware — automating tasks that once required advanced expertise.

Funklocker and SparkCat are early proof points that GenAI-built malware is no longer theoretical, it’s already operational.

SCATTERED SPIDER accelerates identity-based, cross-domain attacks

The group resurged in 2025 with faster and more aggressive tradecraft — leveraging vishing and help desk impersonation to reset credentials, bypass MFA, and move laterally across SaaS and cloud environments.

In one incident, the group moved from initial access to encryption by deploying ransomware in under 24 hours.

China-based adversaries drive continued surge in cloud attacks

Cloud intrusions rose 136%, with China-linked adversaries responsible for 40% of increased activity, as GENESIS PANDA and MURKY PANDA evaded detection through cloud misconfigurations and trusted access.

“The AI era has redefined how businesses operate, and how adversaries attack. We’re seeing threat actors use GenAI to scale social engineering, accelerate operations, and lower the barrier to entry for hands-on-keyboard intrusions,” said Adam Meyers, Head of Counter Adversary Operations at CrowdStrike. “At the same time, adversaries are targeting the very AI systems organisations are deploying. Every AI agent is a superhuman identity: autonomous, fast, and deeply integrated, making them high-value targets.

“Adversaries are treating these agents like infrastructure, attacking them the same way they target SaaS platforms, cloud consoles, and privileged accounts. Securing the AI that powers business is where the cyber battleground is evolving.”

The CrowdStrike 2025 Threat Hunting Report can be downloaded here.

Image credit: iStock.com/Thapana Onphalai