AI-Driven Threat Landscape: Five New Attack Techniques Reshape Cybersecurity in 2025

The cybersecurity landscape is undergoing a seismic shift as the SANS Institute’s recent keynote at the RSA Conference (RSAC) reveals a troubling trend: for the first time in its history, every one of the five most dangerous new attack techniques is intertwined with artificial intelligence (AI). Moderated by Ed Skoudis, President of the SANS Technology Institute, this session serves as a critical early warning system for organizations navigating an increasingly complex threat environment.

Skoudis emphasizes that the intersection of modern infrastructure’s complexity and the rapid deployment of AI by both attackers and defenders creates a dual crisis of speed and comprehension. Organizations must now grapple with the implications of these evolving threats, which are reshaping the cybersecurity landscape in unprecedented ways.

Attack Technique #1: AI-Generated Zero Days, From Scarcity to Surplus

Joshua Wright, Faculty Fellow and Senior Technical Director at SANS Institute, highlights a significant transformation in the realm of zero-day exploits. Historically, the development of zero-day vulnerabilities required extensive research and significant financial investment, often limiting access to well-funded nation-state actors. However, the advent of AI has disrupted this paradigm.

Independent researchers have demonstrated the ability to discover zero-days in widely used software for as little as $116 in AI token costs. This drastic reduction in cost alters the strategic calculus for attackers, making broad exploitation campaigns economically viable for less sophisticated threat actors. Wright notes, “Attackers were already faster than us. AI has made the gap unbridgeable at our current pace.”

The defensive side of this equation is lagging. According to the Verizon 2024 Data Breach Investigations Report (DBIR), half of all critical vulnerabilities remain unpatched for an average of 55 days after a fix is available. This delay becomes untenable when AI can generate new exploits more rapidly than vendors can issue patches. Organizations must enhance their patching processes, automate wherever feasible, and adopt AI-driven detection tools to keep pace with the speed of emerging threats.

Attack Technique #2: Supply Chain Risks, Your Vendor’s Vendor’s Vendor

Wright also addresses the escalating risks associated with supply chain compromises. Recent statistics indicate that two out of three organizations experienced a software supply chain attack in the past year, with third-party involvement in breaches doubling to 30%. In 2025 alone, over 454,000 malicious packages were published to open-source registries, marking a 75% increase from the previous year.

AI-generated patches are enabling malicious actors to produce and distribute compromised code at scale, expanding the attack surface beyond poisoned libraries to include build systems and developer tools. The Shai-Hulud worm, for instance, infected over 1,000 open-source packages, exposing 14,000 credentials across 487 organizations. A China-affiliated group even compromised the Notepad++ update infrastructure for six months, delivering backdoors to targets in critical sectors such as energy and finance.

Wright warns, “Your attack surface is not the software you chose. It is the entire ecosystem of suppliers behind it.” Organizations must proactively plan for potential supplier compromises, demanding verifiable proof of software construction and extending their cybersecurity programs to encompass all update channels and developer tools. Currently, 79% of organizations have cybersecurity measures that cover less than half of their supplier ecosystem, leaving a significant vulnerability that could lead to major compromises.

Attack Technique #3: OT Complexity and the Root Cause Crisis

Robert Lee, SANS Institute Fellow and CEO of Dragos, Inc., sheds light on the complexities surrounding operational technology (OT) incidents. When failures occur within critical infrastructure, understanding the cause is paramount. Recovering systems without grasping the underlying issues can lead to further damage or restore operations in a compromised state.

Lee notes a growing accountability crisis in OT incident response. The evidentiary records of industrial environments—network traffic and commands—are only available if captured prior to an incident. In a case involving a December 2025 attack on Poland’s distributed energy resources, investigators confirmed disruption but could not ascertain the adversary’s actions due to a lack of OT monitoring.

“Governments are not going to be comfortable not knowing what happened in their critical infrastructure and why someone died. That scenario is unacceptable, and it is already happening,” Lee states. The integration of agentic AI into OT environments is compounding these challenges, as many critical infrastructure sectors lack the necessary monitoring infrastructure for effective attribution.

To address these issues, the SANS ICS Five Critical Controls and NERC CIP-015 provide a roadmap for organizations to enhance their security posture. The urgency for investment in these areas cannot wait for the next incident to occur.

Attack Technique #4: The Dark Side of AI, Irresponsible Use in Digital Forensics and Incident Response

Heather Barnhart, Head of Faculty and Senior Forensic Expert at SANS Institute, discusses the risks associated with the irresponsible deployment of AI in digital forensics and incident response (DFIR). While AI can enhance capabilities, its use without proper training and validation can lead to significant failures.

AI systems cannot alert on evidence they are not programmed to recognize, nor can they interpret the significance of missing data as a trained investigator would. In high-stakes investigations, an AI that confidently presents incorrect information without indicating uncertainty poses a liability that can adversely affect case outcomes.

“Most breaches don’t fail because of tools. They fail at decision points. AI cannot be the decision point,” Barnhart asserts. The risks extend beyond investigative accuracy; AI can also be weaponized against organizations through unmonitored channels. For instance, a third-party legal advisor may inadvertently expose sensitive information by using AI services without proper security measures.

The attack surface is not limited to networks; AI acts as a force multiplier, necessitating that trained professionals maintain decision-making authority at every stage of the investigative process.

Attack Technique #5: Find Evil: The Race to Autonomous Defense

Rob T. Lee, Chief AI Officer and Chief of Research at SANS Institute, highlights the rapid evolution of cyberattacks, with AI-driven workflows operating up to 47 times faster than traditional methods. The window for exploiting known vulnerabilities has shrunk dramatically, with attackers able to escalate from stolen credentials to full administrative control of cloud environments in as little as eight minutes.

In November 2025, a campaign known as “GTG 1002,” attributed to a Chinese state-sponsored group, targeted over 30 government and financial organizations. This operation utilized AI tools to automate up to 90% of the attack process, including reconnaissance and lateral movement within networks, often without direct human intervention.

“They have their artificial intelligence. Now we build ours,” Lee emphasizes. This philosophy underpins Protocol SIFT, an open-source initiative from SANS Institute designed to help defenders keep pace with evolving threats. While AI is utilized to streamline workflows and surface insights, human analysts remain responsible for validating findings and making critical decisions.

Early results indicate that this model can significantly reduce response times. In a proof-of-concept exercise involving a complex attack scenario, a Protocol SIFT-assisted analyst completed a full investigation in just over 14 minutes, a task that would typically take several days. This highlights a crucial advantage for defenders: while attackers can scale their tools, they cannot easily replicate the collective coordination of the global security community.

According to publicly available securitymea.com reporting, the implications of these developments are profound, necessitating a reevaluation of how organizations approach cybersecurity in an era dominated by AI-driven threats.

For the latest cybersecurity developments, threat intelligence and breaking updates from across the Middle East: Middle East