04-08-2025
CrowdStrike report warns of GenAI driving surge in cyberattacks
CrowdStrike has released its 2025 Threat Hunting Report detailing how adversaries are using generative AI (GenAI) to enhance and scale cyberattacks, with a particular focus on emerging threats to autonomous AI systems within enterprises.
The report draws on intelligence from CrowdStrike's team of threat hunters and analysts, surveying attacks by over 265 known adversary groups. The findings highlight how attack vectors are evolving with increased automation and use of AI, as well as the targeting of AI-driven systems themselves.
AI-powered attacks
According to the report, GenAI-built malware is now operational, with lower-tier cybercriminals and hacktivist groups utilising AI to generate scripts, troubleshoot technical issues, and develop new forms of malware. Early examples cited include attacks named Funklocker and SparkCat, which underscore how the barrier to entry for sophisticated cybercrime has been lowered.
China-linked adversaries have driven a significant increase in attacks on cloud infrastructure, accounting for 40% of a 136% rise in such incidents during the first half of 2025. The report notes that actors like GENESIS PANDA and MURKY PANDA exploited cloud misconfigurations and access privileges to carry out attacks, while GLACIAL PANDA focused on embedding itself in telecommunications networks, leading to a 130% year-over-year surge in nation-state activity in that sector.
Accelerating social engineering
Beyond technical exploits, the report outlines how AI is being leveraged to automate social engineering campaigns. FAMOUS CHOLLIMA, a North Korea-linked group, used GenAI to generate fraudulent résumés, create deepfake videos for interviews, and complete technical assignments under assumed identities. This group reportedly infiltrated more than 320 companies worldwide, constituting a 220% year-over-year increase. The report also references Russia-linked EMBER BEAR's amplification of pro-Russia narratives and Iran-linked CHARMING KITTEN's deployment of phishing emails crafted with large language models targeting US and EU organisations.
AI agents: A new target
The rise of agentic AI - autonomous AI agents handling key business workflows - has created new opportunities for attackers. Several threat actors have reportedly exploited vulnerabilities in the tools used to build and manage these agents. Access was gained through unauthenticated channels, followed by credential harvesting, malware deployment, and ransomware installation. According to CrowdStrike, this marks the emergence of AI systems, and the identities they use, as a key part of the enterprise attack surface. "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. At the same time, adversaries are targeting the very AI systems organizations 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," said Adam Meyers, Head of Counter Adversary Operations at CrowdStrike.
Trend observations
The report also highlights the resurgence of the SCATTERED SPIDER group, which has accelerated its use of identity-based attacks across multiple domains. The group's tactics in 2025 included using phone-based social engineering (vishing) and impersonation of help desk personnel to reset credentials, bypass multi-factor authentication measures, and deploy ransomware in less than 24 hours after gaining initial access.
CrowdStrike's data shows a clear trend of increased adversary sophistication with the use of AI-enabled tools, not only for direct attacks but also for the exploitation of cloud, SaaS, and AI agent infrastructure. This shift is rapidly transforming both the methods and preferred targets of cybercriminal and nation-state actors.
The report suggests that as enterprises further integrate AI agents into their operations, additional security measures are required to safeguard these autonomous, non-human identities and workflows from being compromised or manipulated.
