April 4, 2022

Explore Internet-Facing System Vulnerabilities

Read about 2021's top four incidents and how Darktrace's advanced threat detection technology identified and mitigated vulnerabilities. Learn more.

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

Apr 2022

By virtue of their exposure, Internet-facing systems (i.e., systems which have ports open/exposed to the wider Internet) are particularly susceptible to compromise. Attackers typically compromise Internet-facing systems by exploiting zero-day vulnerabilities in applications they run. During 2021, critical zero-day vulnerabilities in the following applications were publicly disclosed:

Microsoft Exchange
Confluence
GitLab
Log4j-using applications

Internet-facing systems running these applications were consequently heavily targeted by attackers. In this post, we will provide examples of compromises of these systems observed by Darktrace’s SOC team in 2021. As will become clear, successful exploitation of weaknesses in Internet-facing systems inevitably results in such systems doing things which they do not normally do. Rather than focusing on identifying attempts to exploit these weaknesses, Darktrace focuses on identifying the unusual behaviors which inevitably ensue. The purpose of this post is to highlight the effectiveness of this approach.

Exchange server compromise

In January, researchers from the cyber security company DEVCORE reported a series of critical vulnerabilities in Microsoft Exchange which they dubbed ‘ProxyLogon’.^[1] ProxyLogon consists of a server-side request forgery (SSRF) vulnerability (CVE-2021-26855) and a remote code execution (RCE) vulnerability (CVE-2021-27065). Attackers were observed exploiting these vulnerabilities in the wild from as early as January 6.^[2] In April, DEVCORE researchers reported another series of critical vulnerabilities in Microsoft Exchange which they dubbed ‘ProxyShell’.^[3] ProxyShell consists of a pre-authentication path confusion vulnerability (CVE-2021-34473), a privilege elevation vulnerability (CVE-2021-34523), and a post-authentication RCE vulnerability (CVE-2021-31207). Attackers were first observed exploiting these vulnerabilities in the wild in August.^[4] In many cases, attackers exploited the ProxyShell and ProxyLogon vulnerabilities in order to create web shells on the targeted Exchange servers. The presence of these web shells provided attackers with the means to remotely execute commands on the compromised servers.

In early August 2021, by exploiting the ProxyShell vulnerabilities, an attacker gained the rights to remotely execute PowerShell commands on an Internet-facing Exchange server within the network of a US-based transportation company. The attacker subsequently executed a number of PowerShell commands on the server. One of these commands caused the server to make a 28-minute-long SSL connection to a highly unusual external endpoint. Within a couple of hours, the attacker managed to strengthen their foothold within the network by installing AnyDesk and CobaltStrike on several internal devices. In mid-August, the attacker got the devices on which they had installed Cobalt Strike to conduct network reconnaissance and to transfer terabytes of data to the cloud storage service, MEGA. At the end of August, the attacker got the devices on which they had installed AnyDesk to execute Conti ransomware and to spread executable files and script files to further internal devices.

In this example, the attacker’s exploitation of ProxyShell immediately resulted in the Exchange Server making a long SSL connection to an unusual external endpoint. This connection caused the model Device / Long Agent Connection to New Endpoint to breach. The subsequent reconnaissance, lateral movement, C2, external data transfer, and encryption behavior brought about by the attacker were also picked up by Darktrace’s models.

A non-exhaustive list of the models that breached as a result of the behavior brought about by the attacker:

Device / Long Agent Connection to New Endpoint
Device / ICMP Address Scan
Anomalous Connection / SMB Enumeration
Anomalous Server Activity / Outgoing from Server
Compromise / Beacon to Young Endpoint
Anomalous Server Activity / Rare External from Server
Compromise / Fast Beaconing to DGA
Compromise / SSL or HTTP Beacon
Compromise / Sustained SSL or HTTP Increase
Compromise / Beacon for 4 Days
Anomalous Connection / Multiple HTTP POSTs to Rare Hostname
Unusual Activity / Enhanced Unusual External Data Transfer
Anomalous Connection / Data Sent to Rare Domain
Anomalous Connection / Uncommon 1 GiB Outbound
Compliance / SMB Drive Write
Anomalous File / Internal / Additional Extension Appended to SMB File
Anomalous Connection / Suspicious Read Write Ratio
Anomalous Connection / Suspicious Read Write Ratio and Unusual SMB
Anomalous Connection / Sustained MIME Type Conversion
Unusual Activity / Anomalous SMB Move & Write
Unusual Activity / Unusual Internal Data Volume as Client or Server
Device / Suspicious File Writes to Multiple Hidden SMB Shares
Compromise / Ransomware / Suspicious SMB Activity
Anomalous File / Internal / Unusual SMB Script Write
Anomalous File / Internal / Masqueraded Executable SMB Write
Device / SMB Lateral Movement
Device / Multiple Lateral Movement Model Breaches

Confluence server compromise

Atlassian’s Confluence is an application which provides the means for building collaborative, virtual workspaces. In the era of remote working, the value of such an application is undeniable. The public disclosure of a critical remote code execution (RCE) vulnerability (CVE-2021-26084) in Confluence in August 2021 thus provided a prime opportunity for attackers to cause havoc. The vulnerability, which arises from the use of Object-Graph Navigation Language (OGNL) in Confluence’s tag system, provides attackers with the means to remotely execute code on vulnerable Confluence server by sending a crafted HTTP request containing a malicious parameter.^[5] Attackers were first observed exploiting this vulnerability towards the end of August, and in the majority of cases, attackers exploited the vulnerability in order to install crypto-mining tools onto vulnerable servers.^[6]

At the beginning of September 2021, an attacker was observed exploiting CVE-2021-26084 in order to install the crypto-mining tool, XMRig, as well as a shell script, onto an Internet-facing Confluence server within the network of an EMEA-based television and broadcasting company. Within a couple of hours, the attacker installed files associated with the crypto-mining malware, Kinsing, onto the server. The Kinsing-infected server then immediately began to communicate over HTTP with the attacker’s C2 infrastructure. Around the time of this activity, the server was observed using the MinerGate crypto-mining protocol, indicating that the server had begun to mine cryptocurrency.

In this example, the attacker’s exploitation of CVE-2021-26084 immediately resulted in the Confluence server making an HTTP GET request with an unusual user-agent string (one associated with curl in this case) to a rare external IP. This behavior caused the models Device / New User Agent, Anomalous Connection / New User Agent to IP Without Hostname, and Anomalous File / Script from Rare Location to breach. The subsequent file downloads, C2 traffic and crypto-mining activity also resulted in several models breaching.

A non-exhaustive list of the models which breached as a result of the unusual behavior brought about by the attacker:

Device / New User Agent
Anomalous Connection / New User Agent to IP Without Hostname
Anomalous File / Script from Rare Location
Anomalous File / EXE from Rare External Location
Anomalous File / Internet Facing System File Download
Device / Initial Breach Chain Compromise
Anomalous Connection / Posting HTTP to IP Without Hostname
Compliance / Crypto Currency Mining Activity
Compromise / High Priority Crypto Currency Mining
Device / Internet Facing Device with High Priority Alert

GitLab server compromise

GitLab is an application providing services ranging from project planning to source code management. Back in April 2021, a critical RCE vulnerability (CVE-2021-22205) in GitLab was publicly reported by a cyber security researcher via the bug bounty platform, HackerOne.^[7] The vulnerability, which arises from GitLab’s use of ExifTool for removing metadata from image files, ^[8] enables attackers to remotely execute code on vulnerable GitLab servers by uploading specially crafted image files.^[9] Attackers were first observed exploiting CVE-2021-22205 in the wild in June/July.^[10] A surge in exploitations of the vulnerability was observed at the end of October, with attackers exploiting the flaw in order to assemble botnets.^[11] Darktrace observed a significant number of cases in which attackers exploited the vulnerability in order to install crypto-mining tools onto vulnerable GitLab servers.

On October 29, an attacker successfully exploited CVE-2021-22205 on an Internet-facing GitLab server within the network of a UK-based education provider. The organization was trialing Darktrace when this incident occurred. The attacker installed several executable files and shell scripts onto the server by exploiting the vulnerability. The attacker communicated with the compromised server (using unusual ports) for several days, before making the server transfer large volumes of data externally and download the crypto-mining tool, XMRig, as well as the botnet malware, Mirai. The server was consequently observed making connections to the crypto-mining pool, C3Pool.

In this example, the attacker’s exploitation of the vulnerability in GitLab immediately resulted in the server making an HTTP GET request with an unusual user-agent string (one associated with Wget in this case) to a rare external IP. The models Anomalous Connection / New User Agent to IP Without Hostname and Anomalous File / EXE from Rare External Location breached as a result of this behavior. The attacker’s subsequent activity on the server over the next few days resulted in frequent model breaches.

A non-exhaustive list of the models which breached as a result of the attacker’s activity on the server:

Anomalous Connection / New User Agent to IP Without Hostname
Anomalous File / EXE from Rare External Location
Anomalous File / Multiple EXE from Rare External Locations
Anomalous File / Internet Facing Device with High Priority Alert
Anomalous File / Script from Rare Location
Anomalous Connection / Application Protocol on Uncommon Port
Anomalous Connection / Anomalous SSL without SNI to New External
Device / Initial Breach Chain Compromise
Unusual Activity / Unusual External Data to New IPs
Anomalous Server Activity / Outgoing from Server
Device / Large Number of Model Breaches from Critical Network Device
Anomalous Connection / Data Sent to Rare Domain
Compromise / Suspicious File and C2
Unusual Activity / Enhanced Unusual External Data Transfer
Compliance / Crypto Currency Mining Activity
Compliance / High Priority Crypto Currency Mining
Anomalous File / Zip or Gzip from Rare External Location
Compromise / Monero Mining
Device / Internet Facing Device with High Priority Alert
Anomalous Server Activity / Rare External from Server
Compromise / Slow Beaconing Activity To External Rare
Compromise / Beaconing Activity To External Rare
Compromise / HTTP Beaconing to Rare Destination
Compromise / High Volume of Connections with Beacon Score
Anomalous File / Numeric Exe Download

Log4j server compromise

On December 9 2021, a critical RCE vulnerability (dubbed ‘Log4Shell’) in version 2 of Apache’s Log4j was publicly disclosed by researchers at LunaSec.^[12] As a logging library present in potentially millions of Java applications,^[13] Log4j constitutes an obscured, yet ubiquitous feature of the digital world. The vulnerability (CVE-2021-44228), which arises from Log4j’s Java Naming and Directory Interface (JNDI) Lookup feature, enables an attacker to make a vulnerable server download and execute a malicious Java class file. To exploit the vulnerability, all the attacker must do is submit a specially crafted JNDI lookup request to the server. The fact that Log4j is present in so many applications and that the exploitation of this vulnerability is so simple, Log4Shell has been dubbed the ‘most critical vulnerability of the last decade’.^[14] Attackers have been exploiting Log4Shell in the wild since at least December 1.^[15] Since then, attackers have been observed exploiting the vulnerability to install crypto-mining tools, Cobalt Strike, and RATs onto vulnerable servers.^[16]

On December 10, one day after the public disclosure of Log4Shell, an attacker successfully exploited the vulnerability on a vulnerable Internet-facing server within the network of a US-based architecture company. By exploiting the vulnerability, the attacker managed to get the server to download and execute a Java class file named ‘Exploit69ogQNSQYz.class’. Executing the code in this file caused the server to download a shell script file and a file related to the Kinsing crypto-mining malware. The Kinsing-infected server then went on to communicate over HTTP with a C2 server. Since the customer was using the Proactive Threat Notification (PTN) service, they were immediately alerted to this activity, and the server was subsequently quarantined, preventing crypto-mining activity from taking place.

In this example, the attacker’s exploitation of the zero-day vulnerability immediately resulted in the vulnerable server making an HTTP GET request with an unusual user-agent string (one associated with Java in this case) to a rare external IP. The models Anomalous Connection / Callback on Web Facing Device and Anomalous Connection / New User Agent to IP Without Hostname breached as a result of this behavior. The device’s subsequent file downloads and C2 activity caused several Darktrace models to breach.

A non-exhaustive list of the models which breached as a result of the unusual behavior brought about by the attacker:

Anomalous Connection / Callback on Web Facing Device
Anomalous Connection / New User Agent to IP Without Hostname
Anomalous File / Internet Facing System File Download
Anomalous File / Script from Rare External Location
Device / Initial Breach Chain Compromise
Anomalous Connection / Posting HTTP to IP Without Hostname

Round-up

It is inevitable that attackers will attempt to exploit zero-day vulnerabilities in applications running on Internet-facing devices. Whilst identifying these attempts is useful, the fact that attackers regularly exploit new zero-days makes the task of identifying attempts to exploit them akin to a game of whack-a-mole. Whilst it is uncertain which zero-day vulnerability attackers will exploit next, what is certain is that their exploitation of it will bring about unusual behavior. No matter the vulnerability, whether it be a vulnerability in Microsoft Exchange, Confluence, GitLab, or Log4j, Darktrace will identify the unusual behaviors which inevitably result from its exploitation. By identifying unusual behaviors displayed by Internet-facing devices, Darktrace thus makes it almost impossible for attackers to successfully exploit zero-day vulnerabilities without being detected.

For Darktrace customers who want to find out more about detecting potential compromises of internet-facing devices, refer here for an exclusive supplement to this blog.

Thanks to Andy Lawrence for his contributions.

Footnotes

1. https://devco.re/blog/2021/08/06/a-new-attack-surface-on-MS-exchange-part-1-ProxyLogon/

2. https://www.volexity.com/blog/2021/03/02/active-exploitation-of-microsoft-exchange-zero-day-vulnerabilities/

3. https://www.zerodayinitiative.com/blog/2021/8/17/from-pwn2own-2021-a-new-attack-surface-on-microsoft-exchange-proxyshell

4. https://www.rapid7.com/blog/post/2021/08/12/proxyshell-more-widespread-exploitation-of-microsoft-exchange-servers/

5. https://www.kaspersky.co.uk/blog/confluence-server-cve-2021-26084/23376/

6. https://www.bleepingcomputer.com/news/security/atlassian-confluence-flaw-actively-exploited-to-install-cryptominers/

7. https://hackerone.com/reports/1154542

8. https://security.humanativaspa.it/gitlab-ce-cve-2021-22205-in-the-wild/

9.https://about.gitlab.com/releases/2021/04/14/security-release-gitlab-13-10-3-released/

10. https://www.rapid7.com/blog/post/2021/11/01/gitlab-unauthenticated-remote-code-execution-cve-2021-22205-exploited-in-the-wild/

11. https://www.hackmageddon.com/2021/12/16/1-15-november-2021-cyber-attacks-timeline/

12. https://www.lunasec.io/docs/blog/log4j-zero-day/

13. https://www.csoonline.com/article/3644472/apache-log4j-vulnerability-actively-exploited-impacting-millions-of-java-based-apps.html

14. https://www.theguardian.com/technology/2021/dec/10/software-flaw-most-critical-vulnerability-log-4-shell

15. https://www.rapid7.com/blog/post/2021/12/15/the-everypersons-guide-to-log4shell-cve-2021-44228/

16. https://www.microsoft.com/security/blog/2021/12/11/guidance-for-preventing-detecting-and-hunting-for-cve-2021-44228-log4j-2-exploitation/

Written by

Sam Lister

Specialist Security Researcher

Inside the SOC

Written by

Sam Lister

Specialist Security Researcher

Malware Research Lead

Watch the NIS2 Webinar

Blog

May 28, 2026

From Efficiency to Exposure: How AI Adoption Is Creating Unseen Vulnerabilities on the Factory Floor

How AI agents impact the manufacturing industry

Security teams and IT personnel across the manufacturing industry are under constant pressure to protect production, maintain uptime, and safeguard critical assets but the rise of AI is bringing huge new opportunities alongside new cyber risks. Across manufacturing, AI is embedded into workflows, decision-making, and increasingly, autonomous AI agents are acting on behalf of employees and systems.

Agentic systems are powerful because they can act independently, but that same autonomy also creates cyber and operational risk. Agents have extensive permissions and are capable of carrying out complex tasks, making decisions, and interacting with tools or external systems with little to no human intervention.

Unlike traditional AI models that perform predefined tasks, AI agents use advanced techniques to mimic human decision-making processes, dynamically adapting to new challenges, making decision and taking action based on their own judgement. They look like employees operationally but lack judgment, ethics, or fear of consequences like humans do. This means they can be easily manipulated by cybercriminals, and an AI agent embedded across an OT network creates threats that extend well beyond data exposure. For example, at BMW, AI identifies faults in welding processes as they occur. At its Spartanburg plant, AI monitors the weld of 300-400 metal studs onto every SUV frame to detect misplaced or faulty studs and correct them instantly. Corruption of BMW’s AI system could lead to catastrophic quality control errors.

Adopting agentic AI systems across manufacturing raises some concerns across security teams. New data from our State of AI Cybersecurity survey shows that 78% of manufacturing security professionals are worried about employee use of AI agents – their top concern. That’s followed by employee use of generative AI tools like CoPilot and ChatGPT, a worry for 76% of security professionals at manufacturing organizations. As these tools gain more access to business data and processes, and more autonomy within organizations, security teams, who today have minimal visibility of agent activity in their environments, increasingly have sensitive data exposure (a worry for 60%) and accidental policy and regulatory violations (59%) on their minds.

External AI-powered threats are evolving just as quickly

The same capabilities transforming manufacturing are also reshaping cyberattacks.

AI is enabling attackers to automate reconnaissance, refine targeting, and adapt in real time. What once required time and manual effort can now be executed continuously and at scale. Manufacturers are already seeing the impact. According to manufacturing security professionals we surveyed, 76% are already being impacted by AI-powered threats and 90% see AI increasing the success of social engineering attacks.

And the techniques themselves are evolving. Concerns across the manufacturing sector show growing anxiety about the range of AI-powered attack routes, most pressingly of adaptive malware that evolves in real-time – a prospect half (49%) of manufacturing security professionals we surveyed are worried by, a full 9% more than the average across industries. AI adaptive malware is followed by:

Automated vulnerability scanning and exploit chaining (48%) which has become even more pressing as Anthropic’s new Mythos AI Model supercharges vulnerability discovery

Hyper-personalized phishing campaigns (46%), which remain a mainstay in hackers’ arsenals, and AI has amplified their effectiveness by making phishing emails more convincing and harder to detect.

This is not just an increase in volume, it is a shift toward threats that evolve as they unfold - often faster than static defenses can respond.

Despite rising awareness, many manufacturers are not yet equipped to manage this shift. More than half (51%) say they are not adequately prepared for AI-driven threats, and only 37% have formal policies governing AI deployment.

Securing AI through visibility, context, and guardrails

Addressing this challenge does not require manufacturers to slow innovation. It requires a different approach to security, one that can operate at the same speed and scale as AI. Three specific priorities are emerging for manufacturers looking to take advantage of the power of AI.

Visibility is foundational.

Organizations need to understand where AI is being used, what it can access, and how it behaves across both IT and OT environments. Without that, risk cannot be measured or managed. It is no surprise that Darktrace’s research found that 91% of manufacturing security professionals said that they need to understand how AI makes decisions before trusting it. This is even more critical in operational settings where disruption has safety, environmental, financial, and reputational impacts.

Context is what turns visibility into action.

In environments shaped by AI, normal behavior is constantly shifting. Detecting threats requires a behavioral approach; understanding patterns of life across the organization and identifying subtle deviations in real time – a step change in organizations’ traditional approach to security and risk management.

Guardrails ensure that agency does not become exposure

As AI systems take on greater responsibility, organizations need clear boundaries around what they can do and when they can act independently. These controls must be embedded into systems themselves, not applied after the fact.

Securing AI Agents Across Manufacturing IT and OT

The rise of agentic AI is transforming manufacturing - powering next-generation operations while reshaping the security landscape. This is not just an increase in threats, but a shift to autonomous systems, continuously evolving behaviors, and risks moving at machine speed. For organizations trying to grapple with the challenge of enabling AI while managing the risk, visibility, context and guardrails should be foundational.

Darktrace helps manufacturers build secure AI approaches by making those foundations possible. It provides visibility and real-time detection and response to unusual activity across IT and OT environments and allows organizations to understand AI activity from the prompts employees use and the agents they build to how those agents are behaving across the environment. For manufacturers scaling AI, this delivers a foundation for innovation without sacrificing control.

About the author

Oakley Cox

Director of Product

Blog

May 28, 2026

How to Evaluate AI Vendors: 5 Key categories for AI Adoption

Understanding the AI buyers’ market

AI adoption has become a central topic of discussion in boardrooms, drawing growing interest from business leaders. Ultimately, organizations hope that an investment in AI technology will have tremendous returns. However, the process of buying an AI solution is not as straight forward as it appears on the surface.

While business leaders may be eager to improve productivity across their operations, practitioners responsible for evaluating and selecting AI solutions may not always have the visibility or technical understanding needed to make the right decisions for their business. What is typically marketed as a holistic solution to their most critical problems is usually followed by uncertainty when AI tools are finally operationalized in real environments.

This guide is intended to support security leaders who are under growing pressure to adopt AI tools while navigating complex terminology, vendor claims, and increasingly crowded buying cycles. Ultimately, the goal is to help organizations evaluate and adopt AI in a safe, effective, and well-governed way. To support this, we’ve structured the evaluation framework across five key categories:

Governance, safety, and data controls

Data gathering and training

Model and technique choice

Performance and accuracy validation

Interpretability, adjustability, and transparency

What buying AI looks like in cybersecurity

While investing in AI can bring immense benefits to your security team, first-time buyers of AI cybersecurity solutions may not know where to start. They will have to determine the type of tool they want, know the options available, and evaluate vendors. Research and understanding are critical to ensure purchases are worth the investment.

With acceleration in AI adoption, accompanied by the recent boom in agentic AI and autonomous agents, CISOs must look “beneath the hood" of these tools to understand how they work, how they are governed, and to ensure the system is secure and compliant with internal policies.

Challenges in the AI buyers’ marketplace

The AI security software market is buzzing with hype and flashy promises, which, understandably, needs to be addressed with due diligence. Potential buyers, especially in the cybersecurity space, are hesitant when it comes to allowing AI autonomous capabilities across their workflows, and a lack of vendor transparency can exacerbate those feelings.

Reinforcing this sentiment, research from this year's Darktrace’s State of AI Cybersecurity report shows where confidence and hesitancy emerge amongst potential buyers. On the one hand, security professionals agree that they have good visibility into the logic and reasoning processes their AI solutions use. However, they lack the explainability and trust to allow AI to take independent remedial action.

89% say they have good visibility into the reasoning behind the outputs generated by AI solutions

92% say they need to understand how a defensive AI tool makes decisions before they can trust it

Only 14% say they allow AI to act independently, performing autonomous actions without human approval

74% say they are limiting the autonomy of AI taking action in their SOC until explainability improves

Given the desire for trust and explainability we are seeing from buyers, it's important for them to be equipped with the right questions to ask vendors during an assessment or POV of AI tools in order to demystify marketing hype from real operational outcomes.

Below is a list of categories in which buyers can assess AI vendors or AI Service Providers (AISPs) to help reach safe adoption and maximize their ROI.

5 categories of AI vendor assessment

Darktrace groups these AI-related questions into 5 categories: governance, data and training, model and technique choice, performance validation, and interpretability and adjustability. By asking questions regarding each of these 5 categories, buyers can gain a deeper understanding of how an AISP’s systems work and whether they suit their business requirements.

Governance, safety, and data controls

Governance of AI systems is critical for all AISPs. Whether their platform is based around a single model, or is a more complex, composite AI solution, strong governance is essential to ensure the system is safe, robust, and reliable.

A simple question you could ask is: ‍

What AI governance policies and frameworks do you follow, and/or certifications do you currently maintain?

For more questions you can ask vendors, download the full guide here.

Darktrace is certified to the ISO/IEC 42001 standard, the world’s first AI Management System (AIMS) standard. ISO/IEC 42001 addresses the unique ethical and technical challenges AI poses by setting out a structured way to manage risks such as transparency, accuracy, and misuse. This includes a commitment to ethical AI development, and effective management and monitoring of AI systems both prior to and continually after release.

Data gathering and training

Accurate, meaningful, and unbiased data gathering is the first important step in producing any AI system. An AI model trained using inaccurate, unbalanced, or poor-quality training data will fail to perform optimally.

To alleviate concerns regarding training data quality, a question you could ask is:

‍What steps do you take to prevent bias in your AI models and training data?

For more questions, download the full guide here.

AISPs should be able to provide information about the steps taken, workflows followed, and auditing performed to reduce AI bias where appropriate. While it’s sometimes impossible to fully remove bias from an AI model, appropriate actions should be taken to mitigate or reduce bias where relevant.

Model and technique choice

Different AI techniques are optimal for different tasks. For example, research from Gartner suggests that relying on a single “one-size-fits-all" model can lead to data gaps, especially in highly specialized domains.

To achieve more accurate and robust AI solutions, AI leaders should move beyond using just one model or technique, embrace composite AI practices, and adopt a holistic AI system perspective.

A straightforward question you could ask is simply:

‍What type(s) of AI model(s) do you utilize in your solution?

For more questions, download the full guide here.

While specific detailed information about custom systems used by AISPs is likely proprietary, buyers should expect vendors to be able to provide an overview of the broad techniques used. This will allow you as a buyer to determine if the type of model is appropriate for your use case.

Performance and accuracy validation

Testing and evaluation of performance is essential for all AI systems. Performance analysis should be performed both before release and continually after release to identify potential data or model drift.

A question you could ask to understand an AISPs testing workflow is:

‍How do you audit, test, evaluate, verify, and validate your AI model outputs?

For more questions, download the full guide here.

Testing workflows will likely vary depending on the type of model – measurements relevant to one system may not always be relevant to others. Assessment of systems should also extend beyond these standard accuracy and robustness tests, and should also feature physical performance, such as latency and resource consumption.

Interpretability, adjustability, and transparency

AI systems are typically a black box, simply providing an output without an explanation of how that output was attained. Interpretability and transparency are critical to ensure that both SOC teams and end-users trust the outputs of a system to be accurate and meaningful.

A question you could ask is:

‍How do you promote a trust relationship between human analysts and AI outputs?

For more questions, download the full guide here.

In the context of cybersecurity, trust and interpretability are even more essential. This is particularly relevant for generative AI-based systems (including most AI Agents), where the risk of hallucination can reduce trust in responses.

Cybersecurity systems often need to perform autonomous actions to block incoming threats – an email filtering system may hold potentially dangerous emails; a firewall may block malicious inbound connections. If SOC teams can’t trust these systems to perform accurately, these systems may be limited or disabled, critically reducing their defensive power.

Darktrace as an AI-native cybersecurity vendor

Darktrace has been building and applying AI in cybersecurity for over a decade, developing its capabilities alongside an increasingly complex and fast‑moving threat landscape. This experience has resulted in a mature, multi-layered approach to AI, which continuously learns the normal patterns of each organization to understand behavior, interpret context, and identify meaningful deviations — without relying on predefined rules or known attack signatures. Over time, this has enabled a proven behavioral understanding that helps uncover subtle signals of risk that may otherwise be missed.

With the backing of our ISO/IEC 42001 certification, stakeholders, customers, and partners can be confident that Darktrace is responsibly, ethically, and safely developing its AI systems, and managing the use of AI in day-to-day operations in a compliant and secure manner.

Explore the principles behind Darktrace’s responsible AI approach, informed by collaboration with global experts in academia and governments, detailing how accountability, explainability, and continuous validation are built into its cybersecurity technology.

How Darktrace secures AI systems

Darktrace now brings these capabilities to monitor and respond to risk generated from AI systems across organizations with Darktrace / SECURE AI. This solution analyzes how prompts, agents, and systems are used within the context of each organization, bringing every AI interaction into a single view. This unique approach helps teams understand intent, assess risk, protect sensitive data, and enforce policy across both human and AI agent activity.

Stay up to date

Sign up for the Secure AI Readiness Program here: This gives you exclusive access to the latest news on the latest AI threats, updates on emerging approaches shaping AI security, and insights into the latest innovations, including Darktrace’s ongoing work in this area.

Ready to talk with a Darktrace expert on securing AI? Register here to receive practical guidance on the AI risks that matter most to your business, paired with clarity on where to focus first across governance, visibility, risk reduction, and long-term readiness.