Blog
/
Network
/
June 16, 2025

Tracking CVE-2025-31324: Darktrace’s detection of SAP Netweaver exploitation before and after disclosure 

A critical SAP vulnerability, CVE-2025-31324, allows unauthenticated remote code execution via NetWeaver Visual Composer. Despite early mitigation guidance, many systems remain exposed. Darktrace detected exploitation attempts six days before public disclosure, highlighting the importance of proactive, threat-agnostic detection.
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
Signe Zaharka
Principal Cyber Analyst
person working on laptopDefault blog image
16
Jun 2025

Introduction: Exploiting SAP platforms

Global enterprises depend extensively on SAP platforms, such as SAP NetWeaver and Visual Composer, to run critical business processes worldwide. These systems; however, are increasingly appealing targets for well-resourced adversaries:

What is CVE-2025-31324?

CVE-2025-31324 affects SAP’s NetWeaver Visual Composer, a web-based software modeling tool. SAP NetWeaver is an application server and development platform that runs and connects SAP and non-SAP applications across different technologies [2]. It is commonly used by process specialists to develop application components without coding in government agencies, large enterprises, and by critical infrastructure operators [4].

CVE-2025-31324 affects SAP’s Netweaver Visual Composer Framework 7.1x (all SPS) and above [4]. The vulnerability in a Java Servlet (/irj/servlet_jsp) would enable an unauthorized actor to upload arbitrary files to the /developmentserver/metadatauploader endpoint, potentially resulting in remote code execution (RCE) and full system compromise [3]. The issue stems from an improper authentication and authorization check in the SAP NetWeaver Application Server Java systems [4].

What is the severity rating of CVE-2025-31324?

The vulnerability, first disclosed on April 24, 2025, carries the highest severity rating (CVSS v3 score: 10.0) and could allow remote attackers to upload malicious files without requiring authentication [1][5]. Although SAP released a workaround on April 8, many organizations are hesitant to take their business-critical SAP NetWeaver systems offline, leaving them exposed to potential exploitation [2].

How is CVE-2025-31324 exploited?

The vulnerability is exploitable by sending specifically crafted GET, POST, or HEAD HTTP requests to the /developmentserver/metadatauploader URL using either HTTP or HTTPS. Attackers have been seen uploading malicious files (.jsp, .java, or .class files to paths containing “\irj\servlet_jsp\irj\”), most of them being web shells, to publicly accessible SAP NetWeaver systems.

External researchers observed reconnaissance activity targeting this vulnerability in late January 2025, followed by a surge in exploitation attempts in February. The first confirmed compromise was reported in March [4].

Multiple threat actors have reportedly targeted the vulnerability, including Chinese Advanced Persistent Threats (APTs) groups Chaya_004 [7], UNC5221, UNC5174, and CL-STA-0048 [8], as well as ransomware groups like RansomEXX, also known as Storm-2460, BianLian [4] or Qilin [6] (the latter two share the same indicators of  compromise (IoCs)).

Following the initial workaround published on April 8, SAP released a security update addressing CVE-2025-31324 and subsequently issued a patch on May 13 (Security Note 3604119) to resolve the root cause of the vulnerability [4].

Darktrace’s coverage of CVE-2025-31324 exploitation

Darktrace has observed activity indicative of threat actors exploiting CVE-2025-31324, including one instance detected before the vulnerability was publicly disclosed.

In April 2025, the Darktrace Threat Research team investigated activity related to the CVE-2025-31324 on SAP devices and identified two cases suggesting active exploitation of the vulnerability. One case was detected prior to the public disclosure of the vulnerability, and the other just two days after it was published.

Early detection of CVE 2025-31324 by Darktrace

Figure 1: Timeline of events for an internet-facing system, believed to be a SAP device, exhibiting activity indicative of CVE-2025-31324 exploitation.
Figure 1: Timeline of events for an internet-facing system, believed to be a SAP device, exhibiting activity indicative of CVE-2025-31324 exploitation.

On April 18, six days prior to the public disclosure of CVE-2025-31324, Darktrace began to detect unusual activity on a device belonging to a logistics organization in the Europe, the Middle East and Africa (EMEA) region. Multiple IoCs observed during this incident have since been linked via OSINT to the exploitation of CVE-2025-31324. Notably, however, this reporting was not available at the time of detection, highlighting Darktrace’s ability to detect threats agnostically, without relying on threat intelligence.

The device was observed making  domain name resolution request for the Out-of-Band Application Security Testing (OAST) domain cvvr9gl9namk9u955tsgaxy3upyezhnm6.oast[.]online. OAST is often used by security teams to test if exploitable vulnerabilities exist in a web application but can similarly be used by threat actors for the same purpose [9].

Four days later, on April 22, Darktrace observed the same device, an internet-facing system believed to be a SAP device, downloading multiple executable (.exe) files from several Amazon Simple Storage Service (S3). Darktrace’s Threat Research team later found these files to be associated with the KrustyLoader  malware [23][24][25].

KrustyLoader is known to be associated with the Chinese threat actor UNC5221, also known as UTA0178, which has been reported to aggressively target devices exposed to the internet [10] [14] [15]. It is an initial-stage malware which downloads and launches a second-stage payload – Sliver C2. Sliver is a similar tool to Cobalt Strike (an open-source post-exploitation toolkit). It is used for command-and-control (C2) connections [11][12]13]. After its successful download, KrustyLoader deletes itself to evade detection.  It has been reported that multiple Chinese APT groups have deployed KrustyLoader on SAP Netweaver systems post-compromise [8].

The actors behind KrustyLoader have also been associated with the exploitation of zero-day vulnerabilities in other enterprise systems, including Ivanti devices [12]. Notably, in this case, one of the Amazon S3 domains observed (abode-dashboard-media.s3.ap-south-1.amazonaws[.]com ) had previously been investigated by Darktrace’s Threat Research team as part of their investigation into Ivanti Connect Secure (CS) and Policy Secure (PS) appliances.

In addition to the download of known malicious files, Darktrace also detected new IoCs, including several executable files that could not be attributed to any known malware families or previous attacks, and for which no corresponding OSINT reporting was available.

Figure 2: Darktrace's detection of a likely SAP device downloading an executable file from an Amazon S3 domain on April 22*.

*The model alert was recreated in a demo environment using real incident metadata, as the original customer environment was no longer accessible.

Post-CVE publication detection

Exploit Validation

Between April 27 and 29, Darktrace observed unusual activity from an SAP device on the network of a manufacturing customer in EMEA.

Darktrace / NETWORK’s detection of an SAP device performing a large volume of suspicious activity between April 27 and April 29.
Figure 3: Darktrace / NETWORK’s detection of an SAP device performing a large volume of suspicious activity between April 27 and April 29.

The device was observed making DNS requests for OAST domains (e.g. aaaaaaaa.d06qqn7pu5a6u25tv9q08p5xhbjzw33ge.oast[.]online and aaaaaaaaaaa.d07j2htekalm3139uk2gowmxuhapkijtp.oast[.]pro), suggesting that a threat actor was testing for exploit validation [9].

Darktrace / NETWORK’s detection of a SAP device making suspicious domain name resolution requests for multiple OAST domains.
Figure 4: Darktrace / NETWORK’s detection of a SAP device making suspicious domain name resolution requests for multiple OAST domains.

Privilege escalation tool download attempt

One day later, Darktrace observed the same device attempting to download an executable file from hxxp://23.95.123[.]5:666/xmrigCCall/s.exe (SHA-1 file hash: e007edd4688c5f94a714fee036590a11684d6a3a).

Darktrace / NETWORK identified the user agents Microsoft-CryptoAPI/10.0 and CertUtil URL Agent during the connections to 23.95.123[.]5. The connections were made over port 666, which is not typically used for HTTP connections.

Multiple open-source intelligence (OSINT) vendors have identified the executable file as either JuicyPotato or SweetPotato, both Windows privilege escalation tools[16][17][18][19]. The file hash and the unusual external endpoint have been associated with the Chinese APT group Gelsemium in the past, however, many threat actors are known to leverage this tool in their attacks [20] [21].

Figure 5: Darktrace’s Cyber AI Analyst’s detection of a SAP device downloading a suspicious executable file from hxxp://23.95.123[.]5:666/xmrigCCall/s.exe on April 28, 2025.

Darktrace deemed this activity highly suspicious and triggered an Enhanced Monitoring model alert, a high-priority security model designed to detect activity likely indicative of compromise. As the customer was subscribed to the Managed Threat Detection service, Darktrace’s Security Operations Centre (SOC) promptly investigated the alert and notified the customer for swift remediation. Additionally, Darktrace’s Autonomous Response capability automatically blocked connections to the suspicious IP, 23.95.123[.]5, effectively containing the compromise in its early stages.

Actions taken by Darktrace’s Autonomous Response to block connections to the suspicious external endpoint 23.95.123[.]5. This event log shows that the connections to 23.95.123[.]5 were made over a rare destination port for the HTTP protocol and that new user agents were used during the connections.
Figure 6: Actions taken by Darktrace’s Autonomous Response to block connections to the suspicious external endpoint 23.95.123[.]5. This event log shows that the connections to 23.95.123[.]5 were made over a rare destination port for the HTTP protocol and that new user agents were used during the connections.

Conclusion

The exploitation of CVE-2025-31324 to compromise SAP NetWeaver systems highlights the persistent threat posed by vulnerabilities in public-facing assets. In this case, threat actors leveraged the flaw to gain an initial foothold, followed by attempts to deploy malware linked to groups affiliated with China [8][20].

Crucially, Darktrace demonstrated its ability to detect and respond to emerging threats even before they are publicly disclosed. Six days prior to the public disclosure of CVE-2025-31324, Darktrace detected unusual activity on a device believed to be a SAP system, which ultimately represented an early detection of the CVE. This detection was made possible through Darktrace’s behavioral analysis and anomaly detection, allowing it to recognize unexpected deviations in device behavior without relying on signatures, rules or known IoCs. Combined with its Autonomous Response capability, this allowed for immediate containment of suspicious activity, giving security teams valuable time to investigate and mitigate the threat.

Credit to Signe Zaharka (Principal Cyber Analyst), Emily Megan Lim, (Senior Cyber Analyst) and Ryan Traill (Analyst Content Lead)

Appendices

List of IoCs

23.95.123[.]5:666/xmrigCCall/s.exe - URL- JuicyPotato/SweetPotato - high confidence

29274ca90e6dcf5ae4762739fcbadf01- MD5 file hash - JuicyPotato/SweetPotato - high confidence

e007edd4688c5f94a714fee036590a11684d6a3a - SHA-1 file hash - JuicyPotato/SweetPotato -high confidence

3268f269371a81dbdce8c4eedffd8817c1ec2eadec9ba4ab043cb779c2f8a5d2 - SHA-256 file hash - JuicyPotato/SweetPotato -high confidence

abode-dashboard-media.s3.ap-south-1.amazonaws[.]com/nVW2lsYsYnv58 - URL- high confidence

applr-malbbal.s3.ap-northeast-2.amazonaws[.]com/7p3ow2ZH - URL- high confidence

applr-malbbal.s3.ap-northeast-2.amazonaws[.]com/UUTICMm - URL- KrustyLoader - high confidence

beansdeals-static.s3.amazonaws[.]com/UsjKy - URL- high confidence

brandnav-cms-storage.s3.amazonaws[.]com/3S1kc - URL- KrustyLoader - high confidence

bringthenoiseappnew.s3.amazonaws[.]com/pp79zE - URL- KrustyLoader - high confidence

f662135bdd8bf792a941ea222e8a1330 - MD5 file hash- KrustyLoader - high confidence

fa645f33c0e3a98436a0161b19342f78683dbd9d - SHA-1 file hash- KrustyLoader - high confidence

1d26fff4232bc64f9ab3c2b09281d932dd6afb84a24f32d772d3f7bc23d99c60 - SHA-256 file hash- KrustyLoader - high confidence

6900e844f887321f22dd606a6f2925ef - MD5 file hash- KrustyLoader - high confidence

da23dab4851df3ef7f6e5952a2fc9a6a57ab6983 - SHA-1 file hash- KrustyLoader - high confidence

1544d9392eedf7ae4205dd45ad54ec67e5ce831d2c61875806ce4c86412a4344 - SHA-256 file hash- KrustyLoader - high confidence

83a797e5b47ce6e89440c47f6e33fa08 - MD5 file hash - high confidence

a29e8f030db8990c432020441c91e4b74d4a4e16 - SHA-1 file hash - high confidence

72afde58a1bed7697c0aa7fa8b4e3b03 - MD5 file hash- high confidence

fe931adc0531fd1cb600af0c01f307da3314c5c9 - SHA-1 file hash- high confidence

b8e56de3792dbd0f4239b54cfaad7ece3bd42affa4fbbdd7668492de548b5df8 - SHA-256 file hash- KrustyLoader - high confidence

17d65a9d8d40375b5b939b60f21eb06eb17054fc - SHA-1 file hash- KrustyLoader - high confidence

8c8681e805e0ae7a7d1a609efc000c84 - MD5 file hash- KrustyLoader - high confidence

29274ca90e6dcf5ae4762739fcbadf01 - MD5 file hash- KrustyLoader - high confidence

Darktrace Model Detections

Anomalous Connection / CertUtil Requesting Non Certificate

Anomalous Connection / CertUtil to Rare Destination

Anomalous Connection / Powershell to Rare External

Anomalous File / EXE from Rare External Location

Anomalous File / Multiple EXE from Rare External Locations

Anomalous File / Internet Facing System File Download

Anomalous File / Masqueraded File Transfer (Enhanced Monitoring)

Anomalous Server Activity / New User Agent from Internet Facing System

Compliance / CertUtil External Connection

Compromise / High Priority Tunnelling to Bin Services (Enhanced Monitoring)

Compromise / Possible Tunnelling to Bin Services

Device / Initial Attack Chain Activity (Enhanced Monitoring)

Device / Suspicious Domain

Device / Internet Facing Device with High Priority Alert

Device / Large Number of Model Alerts

Device / Large Number of Model Alerts from Critical Network Device (Enhanced Monitoring)

Device / New PowerShell User Agent

Device / New User Agent

Autonomous Response Model Alerts

Antigena / Network / External Threat / Antigena Suspicious File Block

Antigena / Network / Significant Anomaly / Antigena Controlled and Model Alert

Antigena / Network / Significant Anomaly / Antigena Enhanced Monitoring from Server Block

Antigena / Network / Significant Anomaly / Antigena Significant Server Anomaly Block

Antigena/ Network / External Threat / Antigena Suspicious File Block

Antigena/ Network / External Threat / Antigena Suspicious File Pattern of Life Block

Antigena/ Network / Significant Anomaly / Antigena Alerts Over Time Block

Antigena/ Network / Significant Anomaly / Antigena Controlled and Model Alert

Antigena/ Network / Significant Anomaly / Antigena Enhanced Monitoring from Server Block

Antigena/ Network / Significant Anomaly / Antigena Significant Server Anomaly Block

Cyber AI Analyst Incidents

Possible HTTP Command and Control

Suspicious File Download

MITRE ATT&CK Mapping

Malware - RESOURCE DEVELOPMENT - T1588.001

PowerShell - EXECUTION - T1059.001

Drive-by Compromise - INITIAL ACCESS - T1189

Ingress Tool Transfer - COMMAND AND CONTROL - T1105

Application Layer Protocol - COMMAND AND CONTROL - T1071

Exploitation of Remote Services - LATERAL MOVEMENT - T1210

Exfiltration Over Unencrypted/Obfuscated Non-C2 Protocol - EXFILTRATION - T1048.003

References

1. https://nvd.nist.gov/vuln/detail/CVE-2025-31324

2. https://www.bleepingcomputer.com/news/security/over-1-200-sap-netweaver-servers-vulnerable-to-actively-exploited-flaw/

3. https://reliaquest.com/blog/threat-spotlight-reliaquest-uncovers-vulnerability-behind-sap-netweaver-compromise/

4. https://onapsis.com/blog/active-exploitation-of-sap-vulnerability-cve-2025-31324/

5. https://www.bleepingcomputer.com/news/security/sap-fixes-suspected-netweaver-zero-day-exploited-in-attacks/

6. https://op-c.net/blog/sap-cve-2025-31324-qilin-breach/

7. https://www.forescout.com/blog/threat-analysis-sap-vulnerability-exploited-in-the-wild-by-chinese-threat-actor/

8. https://blog.eclecticiq.com/china-nexus-nation-state-actors-exploit-sap-netweaver-cve-2025-31324-to-target-critical-infrastructures

9. https://portswigger.net/burp/application-security-testing/oast

10. https://www.picussecurity.com/resource/blog/unc5221-cve-2025-22457-ivanti-connect-secure  

11. https://malpedia.caad.fkie.fraunhofer.de/details/elf.krustyloader

12. https://www.broadcom.com/support/security-center/protection-bulletin/krustyloader-backdoor

13. https://labs.withsecure.com/publications/new-krustyloader-variant-dropped-via-screenconnect-exploit

14. https://blog.eclecticiq.com/china-nexus-threat-actor-actively-exploiting-ivanti-endpoint-manager-mobile-cve-2025-4428-vulnerability

15. https://thehackernews.com/2024/01/chinese-hackers-exploiting-critical-vpn.html

16. https://www.virustotal.com/gui/file/3268f269371a81dbdce8c4eedffd8817c1ec2eadec9ba4ab043cb779c2f8a5d2

17. https://bazaar.abuse.ch/sample/3268f269371a81dbdce8c4eedffd8817c1ec2eadec9ba4ab043cb779c2f8a5d2/

18. https://www.fortinet.com/content/dam/fortinet/assets/analyst-reports/report-juicypotato-hacking-tool-discovered.pdf

19. https://www.manageengine.com/log-management/correlation-rules/detecting-sweetpotato.html

20. https://unit42.paloaltonetworks.com/rare-possible-gelsemium-attack-targets-se-asia/

21. https://assets.kpmg.com/content/dam/kpmg/in/pdf/2023/10/kpmg-ctip-gelsemium-apt-31-oct-2023.pdf

22. https://securityaffairs.com/177522/hacking/experts-warn-of-a-second-wave-of-attacks-targeting-sap-netweaver-bug-cve-2025-31324.html

23. https://www.virustotal.com/gui/file/b8e56de3792dbd0f4239b54cfaad7ece3bd42affa4fbbdd7668492de548b5df8

24. https://www.virustotal.com/gui/file/1d26fff4232bc64f9ab3c2b09281d932dd6afb84a24f32d772d3f7bc23d99c60/detection

25. https://www.virustotal.com/gui/file/1544d9392eedf7ae4205dd45ad54ec67e5ce831d2c61875806ce4c86412a4344/detection

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
Signe Zaharka
Principal Cyber Analyst

More in this series

No items found.

Blog

/

AI

/

September 22, 2026

Darktrace / SECURE AI: Extending Behavioral Security for the Age of Agentic AI

Default blog imageDefault blog image

AI is moving faster than the controls built to secure it

Over 80% of Darktrace’s customers now use generative AI services, and the average organization interacted with five different AI provides in August 20261. That level of adoption reflects how quickly AI has become embedded in day-to-day business operations. But as adoption accelerates, so do the opportunities for new forms of exposure that often operate outside traditional security controls. Prompts can expose sensitive information, an over-permissioned agent can turn a routine task into a serious security concern, and an employee using an unsanctioned AI tool to get work done can quietly introduce risk long before the security team is aware of its existence.

Securing AI isn’t a nice-to-have. It's an urgent, board-level requirement for any business that wants AI adoption to be an advantage rather than a liability.

Why traditional security controls fall short

The challenge isn’t a lack of AI security tools, it’s that most security solutions weren’t built for the way AI behaves.  

AI systems are adaptive and, by nature, unpredictable. A prompt can be entirely benign in one context and a serious risk in another. An agent's permissions can look reasonable in isolation and dangerous the moment they're combined with what that agent is actually doing. Rule-based controls, built to catch known signatures and static policy violations, simply aren't designed to catch this kind of subtlety. They might be able to tell you what happened, but they can’t tell you if it mattered.  

Effective AI security requires something different: a deep understanding of what normal looks like across every human, system, and agent in an environment, so the earliest signs of drift, misuse, or compromise stand out as they emerge.

Behavioral understanding is our foundation

Darktrace isn’t reinventing itself to secure AI. We’re extending what we already do. For over a decade, Darktrace has been built on a single premise – that every organization has its own unique, evolving way of operating, shaped by how its systems, users, and devices behave. Understanding that is the only reliable way to catch what rules and signatures miss. Our Adaptive AI™ continuously learns the distinct behaviors, relationships, and operational patterns of every enterprise it protects, building a Unique Behavioral Profile that no other vendor can replicate. It's how we've spent years interpreting ambiguity, uncovering subtle intent, and spotting drift before it becomes dangerous across networks, cloud, identities, email, OT, and endpoints. AI is a new domain, but behavioral security isn't a new discipline for us.

Introducing Darktrace / SECURE AI

Darktrace / SECURE AI helps organizations embrace AI innovation without losing control of how it is used. Delivered through the Darktrace Behavioral Defense Platform™, it provides visibility into AI activity across employees, agents, and AI systems, helping security teams understand how AI is being used throughout their business.

Darktrace then applies behavioral understanding to that activity, adding the context needed to distinguish routine usage from genuine risk. By understanding the relationships between users, agents, systems, and data, / SECURE AI helps organizations move beyond simply observing AI usage to understanding risk, governing behavior against policy, and enabling AI adoption with confidence. The result is greater oversight, stronger governance, and the ability to confidently accelerate AI innovation without creating unmanaged risk.

Securing your AI ecosystem

AI risk doesn't originate from a single source, and effective AI security can't focus on just one layer of the problem. Organizations need visibility, understanding, and governance across the entire AI ecosystem, from the prompts users submit and the agents they create, to the development environments where AI is built and the unsanctioned tools operating outside approved channels. Here's how Darktrace / SECURE AI helps organizations secure each of these areas.

Shadow AI management

Darktrace / SECURE AI helps security teams discover unsanctioned AI services and track usage trends over time using Darktrace telemetry and supported SASE integrations such as Microsoft Entra Global Secure Access. It also extends visibility to Model Context Protocol (MCP) usage, helping teams understand where AI tools and agents are connecting to external services. Through Darktrace platform integrations, organizations can block unsanctioned AI usage or quarantine affected devices when needed, helping them reduce exposure and guide users toward approved options.

AI prompt analysis

Prompts reveal what users are asking AI to do, what information they are sharing, and the outcomes they are trying to produce. Darktrace / SECURE AI provides visibility into prompts, sessions, and responses across supported platforms, including Microsoft Copilot, Copilot Studio, ChatGPT Enterprise, Claude, Salesforce, and AWS Bedrock2. Behavioral analysis detects activity such as attempted jailbreaks, sensitive data exposure, and potential indirect prompt injection, while risk scoring helps analysts prioritize the sessions that need attention. Policy Manager maps organizational AI policies against prompt activity and surfaces potential violations, helping teams govern AI use with direct evidence instead of relying on fragmented logs or inferred intent.

AI agent identities and actions

AI agents have their own permissions, roles, relationships, and access to systems and data. Darktrace / SECURE AI brings agent identities together with the users interacting with them, giving security teams visibility into access, sessions, and connections across supported environments. A real-time audit trail helps teams continuously evaluate whether an agent’s activity remains aligned with its intended purpose, so they can identify excessive access or behavioral drift before it becomes a security issue.

AI agent development risk management

Many AI risks are introduced during development, when agents are created, permissions are assigned, and connections to data sources are established. Darktrace / SECURE AI provides visibility across both low-code and high-code environments. In platforms such as AWS Bedrock, teams can examine AI architecture and the agent identities involved in development and deployment. In low-code environments, such as Copilot Studio, they can observe agents as they are created, monitor whether behavior stays aligned with their intended purpose, and connect prompt activity during development with behavior in production. This helps organizations address misconfigurations and excessive permissions before they reach production.

The promise of AI, secured

The question isn't whether to adopt AI, it's whether security teams can enable secure adoption that allows every business to benefit from AI’s potential.

Darktrace / SECURE AI is built to solve this problem – extending a decade of behavioral understanding to the newest, fastest-moving part of the enterprise, using the same principles which already protect people and hybrid infrastructure.

Darktrace / SECURE AI is generally available now. Discover the product, or get a demo today.

[related-resource]

[1] Based on aggregated Darktrace product telemetry across a fleet of ~8,200 observed deployments, measuring generative AI service usage across accounts monitored in August 2026. The 80%+ figure reflects the share of monitored accounts with any generative AI service usage during the same period.

[2] Microsoft, Copilot, ChatGPT, Claude, Salesforce, and Amazon Bedrock are trademarks of their respective owners.

Continue reading
About the author
Brittany Woodsmall
Senior Product Marketing Manager, AI

Blog

/

AI

/

September 17, 2026

The Problem of Re-defining Human Value in the Agentic Age

Default blog imageDefault blog image

Newsfeeds are constantly informing us about the rapid escalation of agentic AI systems. These systems move far beyond simple machine-based computations, and  the focused, defined and bounded assistance that most AI systems started out as.

The next evolution of AI will harness the agentic properties of orchestration, automation, and heightened value-chains in IT, taking on the burden of workflow management, not just workflow delivery.

In nearly all of these instances, promises are made such as ‘this will free up human time’ or ‘this will allow people to focus on higher-order strategy’. However that message is delivered, one thing is clear: in ceding the orchestration and management of work to increasingly sophisticated AI agents and agentic systems, human value will be elevated to a particular and specific layer: the ability to judge its outputs.  As AI takes on more tasks, humans should be able to focus on a higher level of governance; making sure the decisions that AI offers us are ethical, responsible and worthwhile.  

But there are two major problems with that approach.

This blog discusses the problem of how Agentic systems are re-shaping how we review information, where we fit in, and when we make decisions.  It also discusses the problem of how the increased flow of confident, generated information affects the way we make judgement.  This blog considers how human judgement needs to adapt, and how a behavioral defense approach – using techniques pioneered by Darktrace – can help us do that.

The challenge of knowing where human judgement belongs

As we confer more automated decision-making to agentic systems, it might look increasingly less like ‘granting permissions’, and more like ‘surrendering authority’.  

The judgement layer for AI-generated work is not a fixed boundary. We have become used to the idea of a ‘human in the loop’ (HITL) and, historically, relationships between humans and IT systems were reasonably clear and bounded.  Computer and software systems were programmed to carry out certain tasks or automated functions, and humans could control the gates and decision points where actions were undertaken. Even across highly complex computational workflows, human interaction was a controllable node within the process; we were able to configure and regulate. But in the agentic age, where that human interaction sits, and what it can influence shifts every time AI systems are granted autonomy.  

This leads us to the first problem: if humans are moving themselves (or are being moved) into the ‘judgement’ part of the value chain, exactly where and when do we exercise that judgement?  

Humans are no longer the sole shepherds of computer-based or software-controlled outputs.  We are at times at least one step further (and slower) behind the new agentic shepherds.  We might also be blind to what they are doing.  Not only might we be removed and blind to the actions of our AI shepherds, but with the challenge of unknown, unapproved AI systems operating beyond our control, humans might not even know that our work is being shepherded by an AI at all.  Simply put, with the advent of greater levels of autonomy and orchestration, humans are at risk of not even knowing where to apply our newly-extended powers of strategic judgement.

Shadow AI – the use of unapproved AI systems or processes – is a growing threat to the role of effective governance and oversight. Shadow AI isn't just the AI you can't see. Its the AI you already know about being used in an unapproved way. The ability to generate effective oversight of the AI systems you use (or that are used on your behalf) will be increasingly important to ensure that human judgement in the AI value chain is effective, and deliberately placed.

The problem of what makes good judgement

The second problem lies in how flawed human judgement can be.  Humans are historically, notoriously, and, sometimes dangerously, unreliable when it comes to exercising judgement.  Humans are prone to the worst kinds of bias, the seduction of malign influence, and the sometimes-overwhelming urge to succeed. AI has long had a known flaw of operating with sycophancy, providing outputs that tend to agree with or flatter the human user.  But as AI grows ever more effective, there is a risk of both hyper-enablement (where humans increasingly and knowingly enable AI despite potential harm), as well as the greater risk of suggestion. Both of these aspects could skew the newly-elevated input of human judgement.

Imagine a highly competent AI system that has just orchestrated and managed a dizzying array of processes and workflows.  The AI is designed to present the human decision-maker with recommendations; based on analysis, comparison and other programmed factors.  This is where the human judgement layer is enabled.  But what if that judgement is summarily diffused by an AI-based recommendation that emulates the decision, provides plausible but unattractive alternatives, then suggests (or, worse, directs) the human end-user to take a particular course of action.

The risk here is that you are given a recommendation, tailored to your preferences (which the AI has learned, or which you have divulged), and which appears to make perfect sense.  It appears to be a well-weighted recommendation, with sound arguments that tap into our inherent biases or inclinations so that a specific decision-path is followed. With the growth of agentic systems specifically designed to match user profiles (from Cowork agents to ‘digital twin’ models), the likelihood of agentic influence could badly skew human judgement or, at the least, devalue the proposition that humans are taking a higher-layer of strategic control over AI-based decisions.

If AI convincingly recommends something that may be problematic, it can be difficult to discern both accurate data, and the context required to make the right judgement.

Given the two problems described above, the job of exercising valuable human judgement in the agentic age can draw down to these two questions:

  • When should humans intervene in the agentic process?  
  • How can we make the best possible judgement calls?

What humans contribute that AI cannot

For all the flaws that make human judgement unreliable, people have the edge over even the most sophisticated and powerful AI systems when it comes to issues such as ethics and social context.  An AI system can, with startling granularity, rank the value of adopting a new business proposal: offering predictive metrics on costs, returns, market value, time-to-deliver operations, conformance with legal registers, etc.  But it can’t tell if the business proposal is ethically sound, or if the business venture will potentially affect groups outside of the analyzed proposal. It can’t tell you if the CEO has a ‘bad feeling’ about this effort.  It can’t tell you if this is the right thing to do.  

The ability to add social context, balance complex interpersonal dynamics, understand nuance, and to go beyond what seems economically reasonable is where human judgement can add value.  

Human judgement is difficult to encapsulate in metrics. And the way we train our development may need to adapt too. Rather than building up a gradual, experiential knowledge base, we should think about training the skill of judgement itself; especially for an agentic age.

How behavioral security strengthens AI governance

If this all feels like a vicious circle (‘I need AI help to make good judgements’ / ‘AI can twist what I need to judge’) it needn’t be. The key to this is having a defense-in-depth approach, with tools that can actually help.

This is precisely where behavioral security becomes important. The complex and nuanced way that humans exercise judgement is often rooted in our ability to recognize behavior that doesn't look right. We may not always be able to articulate it immediately, but we can often identify when an action, recommendation, or outcome feels inconsistent with the context around it. As AI systems take on more responsibility across the decision chain, preserving that ability to recognize meaningful deviations becomes increasingly important.

Darktrace’s / SECURE AI is designed to do exactly that. It applies behavioral security to AI ecosystems, helping organizations understand how people, AI tools, identities, and agents interact across the business. By learning the patterns of normal AI usage and surfacing activity that deviates from those patterns, it provides security teams with the context needed to investigate risk, understand unusual behavior, and make informed governance decisions. Rather than relying solely on predefined rules or assumptions, this behavioral understanding helps organizations distinguish between expected AI activity and behavior that warrants closer scrutiny.

This matters because we are already in an era of information overload. If humans are expected to elevate their value through strategic judgement, the ability to do this without being overwhelmed by data (good or bad) will be critical.  

We need the ability to discern when we're being misled by AI, and whether our judgement calls are being made on the basis of accurate, contextual information. Darktrace / SECURE AI provides that additional layer of defensive security for activity we cannot easily see. Whether it is suspected Shadow AI or skewed recommendations, the net result is a protected organization, where users can more effectively use AI to make positive judgements.

For those where that judgement is a critical skill (both individuals, as well as those working in security teams), improving our metacognition - the ability to understand information in a broader context - will supercharge the value of human judgement. When those judgements are grounded in context rather than assumptions we have better information to make sound decisions.

Conclusion

Human judgement is a skill that is honed over time and experience.  Darktrace’s / SECURE AI employs the same principles, but at machine-speed. Rather than influencing or directing, Darktrace / SECURE AI offers AI-enabled assurance; providing human-based judgement with the right context to make a balanced decision.  

What we judge can be valued by the legitimacy of its outputs. For AI, those outputs are valued on the speed and accuracy of the information provided.  Increasingly for humans, the value of our outputs will be based on the validity of our judgement, and how we justify our decisions in ways that engineer confidence.  

Humans often know more than we can express, while AI is prone to expressing more than it truly understands. Humans can bridge the context AI often fails to appreciate. When that judgement is supported by relevant, impartial AI systems, this is the future space where good AI governance will be exercised.

Discover Darktrace / SECURE AI.

[related-resource]

Continue reading
About the author
Jason Lusted
AI Governance Advisor
Your data. Our AI.
Elevate your network security with Darktrace AI