March 12, 2025

Darktrace's Detection of State-Linked ShadowPad Malware

In 2024, Darktrace identified a cluster of intrusions involving the state-linked malware, ShadowPad. This blog will detail ShadowPad and the associated activities detected by Darktrace.

Written by

Sam Lister

Specialist Security Researcher

Inside the SOC

Darktrace cyber analysts are world-class experts in threat intelligence, threat hunting and incident response, and provide 24/7 SOC support to thousands of Darktrace customers around the globe. Inside the SOC is exclusively authored by these experts, providing analysis of cyber incidents and threat trends, based on real-world experience in the field.

Written by

Sam Lister

Specialist Security Researcher

Mar 2025

An integral part of cybersecurity is anomaly detection, which involves identifying unusual patterns or behaviors in network traffic that could indicate malicious activity, such as a cyber-based intrusion. However, attribution remains one of the ever present challenges in cybersecurity. Attribution involves the process of accurately identifying and tracing the source to a specific threat actor(s).

Given the complexity of digital networks and the sophistication of attackers who often use proxies or other methods to disguise their origin, pinpointing the exact source of a cyberattack is an arduous task. Threat actors can use proxy servers, botnets, sophisticated techniques, false flags, etc. Darktrace’s strategy is rooted in the belief that identifying behavioral anomalies is crucial for identifying both known and novel threat actor campaigns.

The ShadowPad cluster

Between July 2024 and November 2024, Darktrace observed a cluster of activity threads sharing notable similarities. The threads began with a malicious actor using compromised user credentials to log in to the target organization's Check Point Remote Access virtual private network (VPN) from an attacker-controlled, remote device named 'DESKTOP-O82ILGG'. In one case, the IP from which the initial login was carried out was observed to be the ExpressVPN IP address, 194.5.83[.]25. After logging in, the actor gained access to service account credentials, likely via exploitation of an information disclosure vulnerability affecting Check Point Security Gateway devices. Recent reporting suggests this could represent exploitation of CVE-2024-24919 [27,28]. The actor then used these compromised service account credentials to move laterally over RDP and SMB, with files related to the modular backdoor, ShadowPad, being delivered to the ‘C:\PerfLogs\’ directory of targeted internal systems. ShadowPad was seen communicating with its command-and-control (C2) infrastructure, 158.247.199[.]185 (dscriy.chtq[.]net), via both HTTPS traffic and DNS tunneling, with subdomains of the domain ‘cybaq.chtq[.]net’ being used in the compromised devices’ TXT DNS queries.

Figure 1: Darktrace’s Advanced Search data showing the VPN-connected device initiating RDP connections to a domain controller (DC). The device subsequently distributes likely ShadowPad-related payloads and makes DRSGetNCChanges requests to a second DC.

Figure 2: Event Log data showing a DC making DNS queries for subdomains of ‘cbaq.chtq[.]net’ to 158.247.199[.]185 after receiving SMB and RDP connections from the VPN-connected device, DESKTOP-O82ILGG.

Darktrace observed these ShadowPad activity threads within the networks of European-based customers in the manufacturing and financial sectors. One of these intrusions was followed a few months later by likely state-sponsored espionage activity, as detailed in the investigation of the year in Darktrace’s Annual Threat Report 2024.

[related-resource]

Related ShadowPad activity

Additional cases of ShadowPad were observed across Darktrace’s customer base in 2024. In some cases, common C2 infrastructure with the cluster discussed above was observed, with dscriy.chtq[.]net and cybaq.chtq[.]net both involved; however, no other common features were identified. These ShadowPad infections were observed between April and November 2024, with customers across multiple regions and sectors affected. Darktrace’s observations align with multiple other public reports that fit the timeframe of this campaign.

Darktrace has also observed other cases of ShadowPad without common infrastructure since September 2024, suggesting the use of this tool by additional threat actors. ‍

The data theft thread

One of the Darktrace customers impacted by the ShadowPad cluster highlighted above was a European manufacturer. A distinct thread of activity occurred within this organization’s network several months after the ShadowPad intrusion, in October 2024.

The thread involved the internal distribution of highly masqueraded executable files via Sever Message Block (SMB) and WMI (Windows Management Instrumentation), the targeted collection of sensitive information from an internal server, and the exfiltration of collected information to a web of likely compromised sites. This observed thread of activity, therefore, consisted of three phrases: lateral movement, collection, and exfiltration.

The lateral movement phase began when an internal user device used an administrative credential to distribute files named ‘ProgramData\Oracle\java.log’ and 'ProgramData\Oracle\duxwfnfo' to the c$ share on another internal system.

*Figure 3: Darktrace model alert highlighting an SMB write of a file named ‘ProgramData\Oracle\java.log’ to the c$ share on another device.*

Over the next few days, Darktrace detected several other internal systems using administrative credentials to upload files with the following names to the c$ share on internal systems:

ProgramData\Adobe\ARM\webservices.dll

ProgramData\Adobe\ARM\wksprt.exe

ProgramData\Oracle\Java\wksprt.exe

ProgramData\Oracle\Java\webservices.dll

ProgramData\Microsoft\DRM\wksprt.exe

ProgramData\Microsoft\DRM\webservices.dll

ProgramData\Abletech\Client\webservices.dll

ProgramData\Abletech\Client\client.exe

ProgramData\Adobe\ARM\rzrmxrwfvp

ProgramData\3Dconnexion\3DxWare\3DxWare.exe

ProgramData\3Dconnexion\3DxWare\webservices.dll

ProgramData\IDMComp\UltraCompare\updater.exe

ProgramData\IDMComp\UltraCompare\webservices.dll

ProgramData\IDMComp\UltraCompare\imtrqjsaqmm

*Figure 4: Cyber AI Analyst highlighting an SMB write of a file named ‘ProgramData\Adobe\ARM\webservices.dll’ to the c$ share on an internal system.*

The threat actor appears to have abused the Microsoft RPC (MS-RPC) service, WMI, to execute distributed payloads, as evidenced by the ExecMethod requests to the IWbemServices RPC interface which immediately followed devices’ SMB uploads.

*Figure 5: Cyber AI Analyst data highlighting a thread of activity starting with an SMB data upload followed by ExecMethod requests.*

Several of the devices involved in these lateral movement activities, both on the source and destination side, were subsequently seen using administrative credentials to download tens of GBs of sensitive data over SMB from a specially selected server. The data gathering stage of the threat sequence indicates that the threat actor had a comprehensive understanding of the organization’s system architecture and had precise objectives for the information they sought to extract.

Immediately after collecting data from the targeted server, devices went on to exfiltrate stolen data to multiple sites. Several other likely compromised sites appear to have been used as general C2 infrastructure for this intrusion activity. The sites used by the threat actor for C2 and data exfiltration purport to be sites for companies offering a variety of service, ranging from consultancy to web design.

Screenshot of one of the likely compromised sites used in the intrusion. — *Figure 6: Screenshot* *of one of the likely compromised sites used in the intrusion.*

At least 16 sites were identified as being likely data exfiltration or C2 sites used by this threat actor in their operation against this organization. The fact that the actor had such a wide web of compromised sites at their disposal suggests that they were well-resourced and highly prepared.

*Figure 7: Darktrace model alert highlighting an internal device slowly exfiltrating data to the external endpoint, yasuconsulting[.]com.*

Figure 8: Darktrace model alert highlighting an internal device downloading nearly 1 GB of data from an internal system just before uploading a similar volume of data to another suspicious endpoint, www.tunemmuhendislik[.]com

Cyber AI Analyst spotlight

*Figure 9: Cyber AI Analyst identifying and piecing together the various steps of a ShadowPad intrusion.*

*Figure 10: Cyber AI Analyst Incident identifying and piecing together the various steps of the data theft activity.*

As shown in the above figures, Cyber AI Analyst’s ability to thread together the different steps of these attack chains are worth highlighting.

In the ShadowPad attack chains, Cyber AI Analyst was able to identify SMB writes from the VPN subnet to the DC, and the C2 connections from the DC. It was also able to weave together this activity into a single thread representing the attacker’s progression.

Similarly, in the data exfiltration attack chain, Cyber AI Analyst identified and connected multiple types of lateral movement over SMB and WMI and external C2 communication to various external endpoints, linking them in a single, connected incident.

These Cyber AI Analyst actions enabled a quicker understanding of the threat actor sequence of events and, in some cases, faster containment.

Attribution puzzle

Publicly shared research into ShadowPad indicates that it is predominantly used as a backdoor in People’s Republic of China (PRC)-sponsored espionage operations [5][6][7][8][9][10]. Most publicly reported intrusions involving ShadowPad are attributed to the China-based threat actor, APT41 [11][12]. Furthermore, Google Threat Intelligence Group (GTIG) recently shared their assessment that ShadowPad usage is restricted to clusters associated with APT41 [13]. Interestingly, however, there have also been public reports of ShadowPad usage in unattributed intrusions [5].

The data theft activity that later occurred in the same Darktrace customer network as one of these ShadowPad compromises appeared to be the targeted collection and exfiltration of sensitive data. Such an objective indicates the activity may have been part of a state-sponsored operation. The tactics, techniques, and procedures (TTPs), artifacts, and C2 infrastructure observed in the data theft thread appear to resemble activity seen in previous Democratic People’s Republic of Korea (DPRK)-linked intrusion activities [15] [16] [17] [18] [19].

The distribution of payloads to the following directory locations appears to be a relatively common behavior in DPRK-sponsored intrusions.

Observed examples:

C:\ProgramData\Oracle\Java\

C:\ProgramData\Adobe\ARM\

C:\ProgramData\Microsoft\DRM\

C:\ProgramData\Abletech\Client\

C:\ProgramData\IDMComp\UltraCompare\

C:\ProgramData\3Dconnexion\3DxWare\

Additionally, the likely compromised websites observed in the data theft thread, along with some of the target URI patterns seen in the C2 communications to these sites, resemble those seen in previously reported DPRK-linked intrusion activities.

No clear evidence was found to link the ShadowPad compromise to the subsequent data theft activity that was observed on the network of the manufacturing customer. It should be noted, however, that no clear signs of initial access were found for the data theft thread – this could suggest the ShadowPad intrusion itself represents the initial point of entry that ultimately led to data exfiltration.

Motivation-wise, it seems plausible for the data theft thread to have been part of a DPRK-sponsored operation. DPRK is known to pursue targets that could potentially fulfil its national security goals and had been publicly reported as being active in months prior to this intrusion [21]. Furthermore, the timing of the data theft aligns with the ratification of the mutual defense treaty between DPRK and Russia and the subsequent accused activities [20].

Darktrace assesses with medium confidence that a nation-state, likely DPRK, was responsible, based on our investigation, the threat actor applied resources, patience, obfuscation, and evasiveness combined with external reporting, collaboration with the cyber community, assessing the attacker’s motivation and world geopolitical timeline, and undisclosed intelligence.

Conclusion

When state-linked cyber activity occurs within an organization’s environment, previously unseen C2 infrastructure and advanced evasion techniques will likely be used. State-linked cyber actors, through their resources and patience, are able to bypass most detection methods, leaving anomaly-based methods as a last line of defense.

Two threads of activity were observed within Darktrace’s customer base over the last year: The first operation involved the abuse of Check Point VPN credentials to log in remotely to organizations’ networks, followed by the distribution of ShadowPad to an internal domain controller. The second operation involved highly targeted data exfiltration from the network of one of the customers impacted by the previously mentioned ShadowPad activity.

Despite definitive attribution remaining unresolved, both the ShadowPad and data exfiltration activities were detected by Darktrace’s Self-Learning AI, with Cyber AI Analyst playing a significant role in identifying and piecing together the various steps of the intrusion activities.

Credit to Sam Lister (R&D Detection Analyst), Emma Foulger (Principal Cyber Analyst), Nathaniel Jones (VP), and the Darktrace Threat Research team.

Appendices

Darktrace / NETWORK model alerts

User / New Admin Credentials on Client

Anomalous Connection / Unusual Admin SMB Session

Compliance / SMB Drive Write

Device / Anomalous SMB Followed By Multiple Model Breaches

Anomalous File / Internal / Unusual SMB Script Write

User / New Admin Credentials on Client

Anomalous Connection / Unusual Admin SMB Session

Compliance / SMB Drive Write

Device / Anomalous SMB Followed By Multiple Model Breaches

Anomalous File / Internal / Unusual SMB Script Write

Device / New or Uncommon WMI Activity

Unusual Activity / Internal Data Transfer

Anomalous Connection / Download and Upload

Anomalous Server Activity / Rare External from Server

Compromise / Beacon to Young Endpoint

Compromise / Agent Beacon (Short Period)

Anomalous Server Activity / Anomalous External Activity from Critical Network Device

Anomalous Connection / POST to PHP on New External Host

Compromise / Sustained SSL or HTTP Increase

Compromise / Sustained TCP Beaconing Activity To Rare Endpoint

Anomalous Connection / Multiple Failed Connections to Rare Endpoint

Device / Multiple C2 Model Alerts

Anomalous Connection / Data Sent to Rare Domain

Anomalous Connection / Download and Upload

Unusual Activity / Unusual External Data Transfer

Anomalous Connection / Low and Slow Exfiltration

Anomalous Connection / Uncommon 1 GiB Outbound

MITRE ATT&CK mapping

(Technique name – Tactic ID)

ShadowPad malware threads

Initial Access - Valid Accounts: Domain Accounts (T1078.002)

Initial Access - External Remote Services (T1133)

Privilege Escalation - Exploitation for Privilege Escalation (T1068)

Privilege Escalation - Valid Accounts: Default Accounts (T1078.001)

Defense Evasion - Masquerading: Match Legitimate Name or Location (T1036.005)

Lateral Movement - Remote Services: Remote Desktop Protocol (T1021.001)

Lateral Movement - Remote Services: SMB/Windows Admin Shares (T1021.002)

Command and Control - Proxy: Internal Proxy (T1090.001)

Command and Control - Application Layer Protocol: Web Protocols (T1071.001)

Command and Control - Encrypted Channel: Asymmetric Cryptography (T1573.002)

Command and Control - Application Layer Protocol: DNS (T1071.004)

Data theft thread

Resource Development - Compromise Infrastructure: Domains (T1584.001)

Privilege Escalation - Valid Accounts: Default Accounts (T1078.001)

Privilege Escalation - Valid Accounts: Domain Accounts (T1078.002)

Execution - Windows Management Instrumentation (T1047)

Defense Evasion - Masquerading: Match Legitimate Name or Location (T1036.005)

Defense Evasion - Obfuscated Files or Information (T1027)

Lateral Movement - Remote Services: SMB/Windows Admin Shares (T1021.002)

Collection - Data from Network Shared Drive (T1039)

Command and Control - Application Layer Protocol: Web Protocols (T1071.001)

Command and Control - Encrypted Channel: Asymmetric Cryptography (T1573.002)

Command and Control - Proxy: External Proxy (T1090.002)

Exfiltration - Exfiltration Over C2 Channel (T1041)

Exfiltration - Data Transfer Size Limits (T1030)

List of indicators of compromise (IoCs)

IP addresses and/or domain names (Mid-high confidence):

ShadowPad thread

- dscriy.chtq[.]net • 158.247.199[.]185 (endpoint of C2 comms)

- cybaq.chtq[.]net (domain name used for DNS tunneling)

Data theft thread

- yasuconsulting[.]com (45.158.12[.]7)

- hobivan[.]net (94.73.151[.]72)

- mediostresbarbas.com[.]ar (75.102.23[.]3)

- mnmathleague[.]org (185.148.129[.]24)

- goldenborek[.]com (94.138.200[.]40)

- tunemmuhendislik[.]com (94.199.206[.]45)

- anvil.org[.]ph (67.209.121[.]137)

- partnerls[.]pl (5.187.53[.]50)

- angoramedikal[.]com (89.19.29[.]128)

- awork-designs[.]dk (78.46.20[.]225)

- digitweco[.]com (38.54.95[.]190)

- duepunti-studio[.]it (89.46.106[.]61)

- scgestor.com[.]br (108.181.92[.]71)

- lacapannadelsilenzio[.]it (86.107.36[.]15)

- lovetamagotchith[.]com (203.170.190[.]137)

- lieta[.]it (78.46.146[.]147)

File names (Mid-high confidence):

ShadowPad thread:

- perflogs\1.txt

- perflogs\AppLaunch.exe

- perflogs\F4A3E8BE.tmp

- perflogs\mscoree.dll

Data theft thread

- ProgramData\Oracle\java.log

- ProgramData\Oracle\duxwfnfo

- ProgramData\Adobe\ARM\webservices.dll

- ProgramData\Adobe\ARM\wksprt.exe

- ProgramData\Oracle\Java\wksprt.exe

- ProgramData\Oracle\Java\webservices.dll

- ProgramData\Microsoft\DRM\wksprt.exe

- ProgramData\Microsoft\DRM\webservices.dll

- ProgramData\Abletech\Client\webservices.dll

- ProgramData\Abletech\Client\client.exe

- ProgramData\Adobe\ARM\rzrmxrwfvp

- ProgramData\3Dconnexion\3DxWare\3DxWare.exe

- ProgramData\3Dconnexion\3DxWare\webservices.dll

- ProgramData\IDMComp\UltraCompare\updater.exe

- ProgramData\IDMComp\UltraCompare\webservices.dll

- ProgramData\IDMComp\UltraCompare\imtrqjsaqmm

- temp\HousecallLauncher64.exe

Attacker-controlled device hostname (Mid-high confidence)

- DESKTOP-O82ILGG

References

[1] https://www.kaspersky.com/about/press-releases/shadowpad-how-attackers-hide-backdoor-in-software-used-by-hundreds-of-large-companies-around-the-world

[2] https://media.kasperskycontenthub.com/wp-content/uploads/sites/43/2017/08/07172148/ShadowPad_technical_description_PDF.pdf

[3] https://blog.avast.com/new-investigations-in-ccleaner-incident-point-to-a-possible-third-stage-that-had-keylogger-capacities

[4] https://securelist.com/operation-shadowhammer-a-high-profile-supply-chain-attack/90380/

[5] https://assets.sentinelone.com/c/Shadowpad?x=P42eqA

[6] https://www.cyfirma.com/research/the-origins-of-apt-41-and-shadowpad-lineage/

[7] https://www.csoonline.com/article/572061/shadowpad-has-become-the-rat-of-choice-for-several-state-sponsored-chinese-apts.html

[8] https://global.ptsecurity.com/analytics/pt-esc-threat-intelligence/shadowpad-new-activity-from-the-winnti-group

[9] https://cymulate.com/threats/shadowpad-privately-sold-malware-espionage-tool/

[10] https://www.secureworks.com/research/shadowpad-malware-analysis

[11] https://blog.talosintelligence.com/chinese-hacking-group-apt41-compromised-taiwanese-government-affiliated-research-institute-with-shadowpad-and-cobaltstrike-2/

[12] https://hackerseye.net/all-blog-items/tails-from-the-shadow-apt-41-injecting-shadowpad-with-sideloading/

[13] https://cloud.google.com/blog/topics/threat-intelligence/scatterbrain-unmasking-poisonplug-obfuscator

[14] https://www.domaintools.com/wp-content/uploads/conceptualizing-a-continuum-of-cyber-threat-attribution.pdf

[15] https://www.nccgroup.com/es/research-blog/north-korea-s-lazarus-their-initial-access-trade-craft-using-social-media-and-social-engineering/

[16] https://www.microsoft.com/en-us/security/blog/2021/01/28/zinc-attacks-against-security-researchers/

[17] https://www.microsoft.com/en-us/security/blog/2022/09/29/zinc-weaponizing-open-source-software/

[18] https://www.welivesecurity.com/en/eset-research/lazarus-luring-employees-trojanized-coding-challenges-case-spanish-aerospace-company/

[19] https://blogs.jpcert.or.jp/en/2021/01/Lazarus_malware2.html

[20] https://usun.usmission.gov/joint-statement-on-the-unlawful-arms-transfer-by-the-democratic-peoples-republic-of-korea-to-russia/

[21] https://media.defense.gov/2024/Jul/25/2003510137/-1/-1/1/Joint-CSA-North-Korea-Cyber-Espionage-Advance-Military-Nuclear-Programs.PDF

[22] https://kyivindependent.com/first-north-korean-troops-deployed-to-front-line-in-kursk-oblast-ukraines-military-intelligence-says/

[23] https://www.microsoft.com/en-us/security/blog/2024/12/04/frequent-freeloader-part-i-secret-blizzard-compromising-storm-0156-infrastructure-for-espionage/

[24] https://www.microsoft.com/en-us/security/blog/2024/12/11/frequent-freeloader-part-ii-russian-actor-secret-blizzard-using-tools-of-other-groups-to-attack-ukraine/

[25] https://www.sentinelone.com/labs/chamelgang-attacking-critical-infrastructure-with-ransomware/

[26] https://thehackernews.com/2022/06/state-backed-hackers-using-ransomware.html/

[27] https://blog.checkpoint.com/security/check-point-research-explains-shadow-pad-nailaolocker-and-its-protection/

[28] https://www.orangecyberdefense.com/global/blog/cert-news/meet-nailaolocker-a-ransomware-distributed-in-europe-by-shadowpad-and-plugx-backdoors

[related-resource]

AI Cybersecurity: Insights for 2025

We surveyed 1,500+ cybersecurity professionals globally to explore their views, knowledge, and priorities on AI cybersecurity in 2025.

Download now

Written by

Sam Lister

Specialist Security Researcher

Inside the SOC

Written by

Sam Lister

Specialist Security Researcher

Product Marketing Manager, AI & Attack Surface

Watch the NIS2 Webinar

Blog

AI

December 23, 2025

How to Secure AI in the Enterprise: A Practical Framework for Models, Data, and Agents

Introduction: Why securing AI is now a security priority

AI adoption is at the forefront of the digital movement in businesses, outpacing the rate at which IT and security professionals can set up governance models and security parameters. Adopting Generative AI chatbots, autonomous agents, and AI-enabled SaaS tools promises efficiency and speed but also introduces new forms of risk that traditional security controls were never designed to manage. For many organizations, the first challenge is not whether AI should be secured, but what “securing AI” actually means in practice. Is it about protecting models? Governing data? Monitoring outputs? Or controlling how AI agents behave once deployed?

While demand for adoption increases, securing AI use in the enterprise is still an abstract concept to many and operationalizing its use goes far beyond just having visibility. Practitioners need to also consider how AI is sourced, built, deployed, used, and governed across the enterprise.

The goal for security teams: Implement a clear, lifecycle-based AI security framework. This blog will demonstrate the variety of AI use cases that should be considered when developing this framework and how to frame this conversation to non-technical audiences.

What does “securing AI” actually mean?

Securing AI is often framed as an extension of existing security disciplines. In practice, this assumption can cause confusion.

Traditional security functions are built around relatively stable boundaries. Application security focuses on code and logic. Cloud security governs infrastructure and identity. Data security protects sensitive information at rest and in motion. Identity security controls who can access systems and services. Each function has clear ownership, established tooling, and well-understood failure modes.

AI does not fit neatly into any of these categories. An AI system is simultaneously:

An application that executes logic
A data processor that ingests and generates sensitive information
A decision-making layer that influences or automates actions
A dynamic system that changes behavior over time

As a result, the security risks introduced by AI cuts across multiple domains at once. A single AI interaction can involve identity misuse, data exposure, application logic abuse, and supply chain risk all within the same workflow. This is where the traditional lines between security functions begin to blur.

For example, a malicious prompt submitted by an authorized user is not a classic identity breach, yet it can trigger data leakage or unauthorized actions. An AI agent calling an external service may appear as legitimate application behavior, even as it violates data sovereignty or compliance requirements. AI-generated code may pass standard development checks while introducing subtle vulnerabilities or compromised dependencies.

In each case, no single security team “owns” the risk outright.

This is why securing AI cannot be reduced to model safety, governance policies, or perimeter controls alone. It requires a shared security lens that spans development, operations, data handling, and user interaction. Securing AI means understanding not just whether systems are accessed securely, but whether they are being used, trained, and allowed to act in ways that align with business intent and risk tolerance.

At its core, securing AI is about restoring clarity in environments where accountability can quickly blur. It is about knowing where AI exists, how it behaves, what it is allowed to do, and how its decisions affect the wider enterprise. Without this clarity, AI becomes a force multiplier for both productivity and risk.

The five categories of AI risk in the enterprise

A practical way to approach AI security is to organize risk around how AI is used and where it operates. The framework below defines five categories of AI risk, each aligned to a distinct layer of the enterprise AI ecosystem

How to Secure AI in the Enterprise:

Defending against misuse and emergent behaviors
Monitoring and controlling AI in operation
Protecting AI development and infrastructure
Securing the AI supply chain
Strengthening readiness and oversight

Together, these categories provide a structured lens for understanding how AI risk manifests and where security teams should focus their efforts.

1. Defending against misuse and emergent AI behaviors

Generative AI systems and agents can be manipulated in ways that bypass traditional controls. Even when access is authorized, AI can be misused, repurposed, or influenced through carefully crafted prompts and interactions.

Key risks include:

Malicious prompt injection designed to coerce unwanted actions
Unauthorized or unintended use cases that bypass guardrails
Exposure of sensitive data through prompt histories
Hallucinated or malicious outputs that influence human behavior

Unlike traditional applications, AI systems can produce harmful outcomes without being explicitly compromised. Securing this layer requires monitoring intent, not just access. Security teams need visibility into how AI systems are being prompted, how outputs are consumed, and whether usage aligns with approved business purposes

2. Monitoring and controlling AI in operation

Once deployed, AI agents operate at machine speed and scale. They can initiate actions, exchange data, and interact with other systems with little human oversight. This makes runtime visibility critical.

Operational AI risks include:

Agents using permissions in unintended ways
Uncontrolled outbound connections to external services or agents
Loss of forensic visibility into ephemeral AI components
Non-compliant data transmission across jurisdictions

Securing AI in operation requires real-time monitoring of agent behavior, centralized control points such as AI gateways, and the ability to capture agent state for investigation. Without these capabilities, security teams may be blind to how AI systems behave once live, particularly in cloud-native or regulated environments.

3. Protecting AI development and infrastructure

Many AI risks are introduced long before deployment. Development pipelines, infrastructure configurations, and architectural decisions all influence the security posture of AI systems.

Common risks include:

Misconfigured permissions and guardrails
Insecure or overly complex agent architectures
Infrastructure-as-Code introducing silent misconfigurations
Vulnerabilities in AI-generated code and dependencies

AI-generated code adds a new dimension of risk, as hallucinated packages or insecure logic may be harder to detect and debug than human-written code. Securing AI development means applying security controls early, including static analysis, architectural review, and continuous configuration monitoring throughout the build process.

4. Securing the AI supply chain

AI supply chains are often opaque. Models, datasets, dependencies, and services may come from third parties with varying levels of transparency and assurance.

Key supply chain risks include:

Shadow AI tools used outside approved controls
External AI agents granted internal access
Suppliers applying AI to enterprise data without disclosure
Compromised models, training data, or dependencies

Securing the AI supply chain requires discovering where AI is used, validating the provenance and licensing of models and data, and assessing how suppliers process and protect enterprise information. Without this visibility, organizations risk data leakage, regulatory exposure, and downstream compromise through trusted integrations.

5. Strengthening readiness and oversight

Even with strong technical controls, AI security fails without governance, testing, and trained teams. AI introduces new incident scenarios that many security teams are not yet prepared to handle.

Oversight risks include:

Lack of meaningful AI risk reporting
Untested AI systems in production
Security teams untrained in AI-specific threats

Organizations need AI-aware reporting, red and purple team exercises that include AI systems, and ongoing training to build operational readiness. These capabilities ensure AI risks are understood, tested, and continuously improved, rather than discovered during a live incident.

Reframing AI security for the boardroom

AI security is not just a technical issue. It is a trust, accountability, and resilience issue. Boards want assurance that AI-driven decisions are reliable, explainable, and protected from tampering.

Effective communication with leadership focuses on:

Trust: confidence in data integrity, model behavior, and outputs
Accountability: clear ownership across teams and suppliers
Resilience: the ability to operate, audit, and adapt under attack or regulation

Mapping AI security efforts to recognized frameworks such as ISO/IEC 42001 and the NIST AI Risk Management Framework helps demonstrate maturity and aligns AI security with broader governance objectives.

Conclusion: Securing AI is a lifecycle challenge

The same characteristics that make AI transformative also make it difficult to secure. AI systems blur traditional boundaries between software, users, and decision-making, expanding the attack surface in subtle but significant ways.

Securing AI requires restoring clarity. Knowing where AI exists, how it behaves, who controls it, and how it is governed. A framework-based approach allows organizations to innovate with AI while maintaining trust, accountability, and control.

The journey to secure AI is ongoing, but it begins with understanding the risks across the full AI lifecycle and building security practices that evolve alongside the technology.

About the author

Brittany Woodsmall

Product Marketing Manager, AI & Attack Surface

Blog

AI

December 22, 2025

The Year Ahead: AI Cybersecurity Trends to Watch in 2026

Introduction: 2026 cyber trends

Each year, we ask some of our experts to step back from the day-to-day pace of incidents, vulnerabilities, and headlines to reflect on the forces reshaping the threat landscape. The goal is simple: to identify and share the trends we believe will matter most in the year ahead, based on the real-world challenges our customers are facing, the technology and issues our R&D teams are exploring, and our observations of how both attackers and defenders are adapting.

In 2025, we saw generative AI and early agentic systems moving from limited pilots into more widespread adoption across enterprises. Generative AI tools became embedded in SaaS products and enterprise workflows we rely on every day, AI agents gained more access to data and systems, and we saw glimpses of how threat actors can manipulate commercial AI models for attacks. At the same time, expanding cloud and SaaS ecosystems and the increasing use of automation continued to stretch traditional security assumptions.

Looking ahead to 2026, we’re already seeing the security of AI models, agents, and the identities that power them becoming a key point of tension – and opportunity -- for both attackers and defenders. Long-standing challenges and risks such as identity, trust, data integrity, and human decision-making will not disappear, but AI and automation will increase the speed and scale of the cyber risk.

Here's what a few of our experts believe are the trends that will shape this next phase of cybersecurity, and the realities organizations should prepare for.

Agentic AI is the next big insider risk

‍In 2026, organizations may experience their first large-scale security incidents driven by agentic AI behaving in unintended ways—not necessarily due to malicious intent, but because of how easily agents can be influenced. AI agents are designed to be helpful, lack judgment, and operate without understanding context or consequence. This makes them highly efficient—and highly pliable. Unlike human insiders, agentic systems do not need to be socially engineered, coerced, or bribed. They only need to be prompted creatively, misinterpret legitimate prompts, or be vulnerable to indirect prompt injection. Without strong controls around access, scope, and behavior, agents may over-share data, misroute communications, or take actions that introduce real business risk. Securing AI adoption will increasingly depend on treating agents as first-class identities—monitored, constrained, and evaluated based on behavior, not intent.

-- Nicole Carignan, SVP of Security & AI Strategy ‍

Prompt Injection moves from theory to front-page breach

‍We’ll see the first major story of an indirect prompt injection attack against companies adopting AI either through an accessible chatbot or an agentic system ingesting a hidden prompt. In practice, this may result in unauthorized data exposure or unintended malicious behavior by AI systems, such as over-sharing information, misrouting communications, or acting outside their intended scope. Recent attention on this risk—particularly in the context of AI-powered browsers and additional safety layers being introduced to guide agent behavior—highlights a growing industry awareness of the challenge.

-- Collin Chapleau, Senior Director of Security & AI Strategy ‍

Humans are even more outpaced, but not broken

‍When it comes to cyber, people aren’t failing; the system is moving faster than they can. Attackers exploit the gap between human judgment and machine-speed operations. The rise of deepfakes and emotion-driven scams that we’ve seen in the last few years reduce our ability to spot the familiar human cues we’ve been taught to look out for. Fraud now spans social platforms, encrypted chat, and instant payments in minutes. Expecting humans to be the last line of defense is unrealistic.

Defense must assume human fallibility and design accordingly. Automated provenance checks, cryptographic signatures, and dual-channel verification should precede human judgment. Training still matters, but it cannot close the gap alone. In the year ahead, we need to see more of a focus on partnership: systems that absorb risk so humans make decisions in context, not under pressure.

-- Margaret Cunningham, VP of Security & AI Strategy

‍AI removes the attacker bottleneck—smaller organizations feel the impact

‍One factor that is currently preventing more companies from breaches is a bottleneck on the attacker side: there’s not enough human hacker capital. The number of human hands on a keyboard is a rate-determining factor in the threat landscape. Further advancements of AI and automation will continue to open that bottleneck. We are already seeing that. The ostrich approach of hoping that one’s own company is too obscure to be noticed by attackers will no longer work as attacker capacity increases.

-- Max Heinemeyer, Global Field CISO ‍

SaaS platforms become the preferred supply chain target

‍Attackers have learned a simple lesson: compromising SaaS platforms can have big payouts. As a result, we’ll see more targeting of commercial off-the-shelf SaaS providers, which are often highly trusted and deeply integrated into business environments. Some of these attacks may involve software with unfamiliar brand names, but their downstream impact will be significant. In 2026, expect more breaches where attackers leverage valid credentials, APIs, or misconfigurations to bypass traditional defenses entirely.

-- Nathaniel Jones, VP of Security & AI Strategy ‍

Increased commercialization of generative AI and AI assistants in cyber attacks

One trend we’re watching closely for 2026 is the commercialization of AI-assisted cybercrime. For example, cybercrime prompt playbooks sold on the dark web—essentially copy-and-paste frameworks that show attackers how to misuse or jailbreak AI models. It’s an evolution of what we saw in 2025, where AI lowered the barrier to entry. In 2026, those techniques become productized, scalable, and much easier to reuse.

-- Toby Lewis, Global Head of Threat Analysis

Conclusion

Taken together, these trends underscore that the core challenges of cybersecurity are not changing dramatically -- identity, trust, data, and human decision-making still sit at the core of most incidents. What is changing quickly is the environment in which these challenges play out. AI and automation are accelerating everything: how quickly attackers can scale, how widely risk is distributed, and how easily unintended behavior can create real impact. And as technology like cloud services and SaaS platforms become even more deeply integrated into businesses, the potential attack surface continues to expand.

Predictions are not guarantees. But the patterns emerging today suggest that 2026 will be a year where securing AI becomes inseparable from securing the business itself. The organizations that prepare now—by understanding how AI is used, how it behaves, and how it can be misused—will be best positioned to adopt these technologies with confidence in the year ahead.

Learn more about how to secure AI adoption in the enterprise without compromise by registering to join our live launch webinar on February 3, 2026.

About the author

The Darktrace Community

Your data. Our AI.

Elevate your network security with Darktrace AI

Get a demo

Check out this article by Darktrace: Darktrace's Detection of State-Linked ShadowPad Malware

Darktrace's Detection of State-Linked ShadowPad Malware

The ShadowPad cluster

Related ShadowPad activity

The data theft thread

Cyber AI Analyst spotlight

Attribution puzzle

Conclusion

Appendices

MITRE ATT&CK mapping

ShadowPad malware threads

Data theft thread

List of indicators of compromise (IoCs)

File names (Mid-high confidence):

References

AI Cybersecurity: Insights for 2025

Beyond MFA: Detecting Adversary-in-the-Middle Attacks and Phishing with Darktrace

How to Secure AI in the Enterprise: A Practical Framework for Models, Data, and Agents

Enjoying the blog?

More in this series

React2Shell: How Opportunist Attackers Exploited CVE-2025-55182 Within Hours

Atomic Stealer: Darktrace’s Investigation of a Growing macOS Threat

CastleLoader & CastleRAT: Behind TAG150’s Modular Malware Delivery System

Blog

AI

December 23, 2025

How to Secure AI in the Enterprise: A Practical Framework for Models, Data, and Agents

Introduction: Why securing AI is now a security priority

What does “securing AI” actually mean?

The five categories of AI risk in the enterprise

1. Defending against misuse and emergent AI behaviors

2. Monitoring and controlling AI in operation

3. Protecting AI development and infrastructure

4. Securing the AI supply chain

5. Strengthening readiness and oversight

Reframing AI security for the boardroom

Conclusion: Securing AI is a lifecycle challenge

Blog

AI

December 22, 2025

The Year Ahead: AI Cybersecurity Trends to Watch in 2026

Introduction: 2026 cyber trends

Agentic AI is the next big insider risk

Prompt Injection moves from theory to front-page breach

Humans are even more outpaced, but not broken

‍AI removes the attacker bottleneck—smaller organizations feel the impact

SaaS platforms become the preferred supply chain target

Increased commercialization of generative AI and AI assistants in cyber attacks

Conclusion