Blog
/
/
August 10, 2022

Cyber Tactics in the Russo-Ukrainian Conflict

The conflict between Russia and Ukraine has led to fears of a full-scale cyberwar. Learn the cyber attack tactics used, hacking groups involved, and more!
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
Rosa Jong
OSINT Analyst
Written by
Taisiia Garkava
Security Analyst
Default blog image
10
Aug 2022

Introduction

Since the beginning of the Russian invasion of Ukraine in February 2022, cyber communities around the world have been witnessing what can be called a ‘renaissance of cyberwarfare' [1]. Rather than being financially motivated, threat actors are being guided by political convictions to defend allies or attack their enemies. This blog reviews some of the main threat actors involved in this conflict and their ongoing tactics, and advises on how organizations can best protect themselves. Darktrace’s preliminary assessments predicted that attacks would be observed globally with a focus on pro-Ukrainian nations such as North Atlantic Treaty Organization (NATO) members and that identified Advanced Persistent Threat (APT) groups would develop new and complex malware deployed through increasingly sophisticated attack vectors. This blog will show that many of these assessments had unexpected outcomes.

Context for Conflict 

Cyber confrontation between Russia and Ukraine dates back to 2013, when Viktor Yanukovych, (former President of Ukraine) rejected an EU trade pact in favour of an agreement with Russia. This sparked mass protests leading to his overthrow, and shortly after, Russian troops annexed Crimea and initiated the beginning of Russian-Ukrainian ground and cyber warfare. Since then, Russian threat actors have been periodically targeting Ukrainian infrastructure. One of the most notable examples of this, an attack against their national power grid in December 2015, resulted in power outages for approximately 255,000 people in Ukraine and was later attributed to the Russian hacking group Sandworm [2 & 3]. 

Another well-known attack in June 2017 overwhelmed the websites of hundreds of Ukrainian organizations using the infamous NotPetya malware. This attack is still considered the most damaging cyberattack in history, with more than €10 billion euros in financial damage [4]. In February 2022, countries witnessed the next stage of cyberwar against Ukraine with both new and familiar actors deploying various techniques to target their rival’s critical infrastructure. 

Tactic 1: Ransomware

Although some sources suggest US ransomware incidents and expectations of ransom may have declined during the conflict, ransomware still remained a significant tactic deployed globally across this period [5] [6] [7]. A Ukrainian hacking group, Network Battalion 65 (NB65), used ransomware to attack the Russian state-owned television and radio broadcasting network VGTRK. NB65 managed to steal 900,000 emails and 4000 files, and later demanded a ransom which they promised to donate to the Ukrainian army. This attack was unique because the group used the previously leaked source code of Conti, another infamous hacker group that had pledged its support to the Russian government earlier in the conflict. NB65 modified the leaked code to make unique ransomware for each of its targets [5]. 

Against expectations, Darktrace’s customer base appeared to deviate from these ransom trends. Analysts have seen relatively unsophisticated ransomware attacks during the conflict period, with limited evidence to suggest they were connected to any APT activity. Between November 2021 and June 2022, there were 51 confirmed ransomware compromises across the Darktrace customer base. This represents an increase of 43.16% compared to the same period the year before, accounting for relative customer growth. Whilst this suggests an overall growth in ransom cases, many of these confirmed incidents were unattributed and did not appear to be targeting any particular verticals or regions. While there was an increase in the energy sector, this could not be explicitly linked to the conflict. 

The Darktrace DETECT family has a variety of models related to ransomware visibility:

Darktrace Detections for T1486 (Data Encrypted for Impact):

- Compromise / Ransomware / Ransom or Offensive Words Written to SMB

- Compromise / Ransomware / Suspicious SMB Activity

- Anomalous Connection / Sustained MIME Type Conversion

- Unusual Activity / Sustained Anomalous SMB Activity

- Compromise / Ransomware / Suspicious SMB File Extension

- Unusual Activity / Anomalous SMB Read & Write

- Unusual Activity / Anomalous SMB Read & Write from New Device

- SaaS / Resource / SaaS Resources with Additional Extensions

- Compromise / Ransomware / Possible Ransom Note Read

- [If RESPOND is enabled] Antigena / Network / External Threat / Antigena Ransomware Block

Tactic 2: Wipers

One of the largest groups of executables seen during the conflict were wipers. On the eve of the invasion, Ukrainian organizations were targeted by a new wiper malware given the name “HermeticWiper”. Hermetic refers to the name of the Cyprian company “Hermetica Digital Ltd.” which was used by attackers to request a code signing certificate [6]. Such a digital certificate is used to verify the ownership of the code and that it has not been altered. The 24-year-old owner of Hermetica Digital says he had no idea that his company was abused to retrieve a code signing certificate [7]. 

HermeticWiper consists of three components: a worm, decoy ransomware and the wiper malware. The custom worm designed for HermeticWiper was used to spread the malware across the network of its infected machines. ESET researchers discovered that the decoy ransomware and the wiper were released at the same time [8]. The decoy ransomware was used to make it look like the machine was hit by ransomware, when in reality the wiper was already permanently wiping data from the machines. In the attack’s initial stage, it bypasses Windows security features designed to prevent overwriting boot records by installing a separate driver. After wiping data from the machine, HermeticWiper prevents that data from being re-fragmented and overwrites the files to fragment it further. This is done to make it more challenging to reconstruct data for post-compromise forensics [9]. Overall, the function and purpose of HermeticWiper seems similar to that of NotPetya ransomware. 

HermeticWiper is not the only conflict-associated wiper malware which has been observed. In January 2022, Microsoft warned Ukrainian customers that they detected wiper intrusion activity against several European organizations. One example of this was the MBR (Master Boot Record) wiper. This type of wiper overwrites the MBR, the disk sector that instructs a computer on how to load its operating system, with a ransomware note. In reality, the note is a misdirection and the malware destroys the MBR and targeted files [10].  

One of the most notable groups that used wiper malware was Sandworm. Sandworm is an APT attributed to Russia’s foreign military intelligence agency, GRU. The group has been active since 2009 and has used a variety of TTPs within their attacks. They have a history of targeting Ukraine including attacks in 2015 on Ukraine’s energy distribution companies and in 2017 when they used the aforementioned NotPetya malware against several Ukrainian organizations [11]. Another Russian (or pro-Russian) group using wiper malware to target Ukraine is DEV-0586. This group targeted various Ukrainian organizations in January 2022 with Whispergate wiper malware. This type of wiper malware presents itself as ransomware by displaying a file instructing the victim to pay Bitcoin to have their files decrypted [12].  

Darktrace did not observe any confirmed cases of HermeticWiper nor other conflict-associated wipers (e.g IsaacWiper and CaddyWiper) within the customer base over this period. Despite this, Darktrace DETECT has a variety of models related to wipers and data destruction:

Darktrace Detections for T1485 (Data Destruction)- this is the main technique exploited during wiper attacks

- Unusual Activity / Anomalous SMB Delete Volume

- IaaS / Unusual Activity / Anomalous AWS Resources Deleted

- IaaS / Storage / S3 Bucket Delete

- SaaS / Resource / Mass Email Deletes from Rare Location

- SaaS / Resource / Anomalous SaaS Resources Deleted

- SaaS / Resource / Resource Permanent Delete

- [If RESPOND is enabled] Antigena / Network / Manual / Enforce Pattern of Life

- [If RESPOND is enabled] Antigena / SaaS / Antigena Unusual Activity Block

Tactic 3: Spear-Phishing

Another strategy that some threat actors employ is spear-phishing. Targeting can be done using email, social media, messaging, or other platforms.

The hacking group Armageddon (also known as Gamaredon) has been responsible for several spear-phishing attacks during the crisis, primarily targeting individuals involved in the Ukrainian Government [13]. Since the beginning of the war, the group has been sending out a large volume of emails containing an HTML file which, if opened, downloads and launches a RAR payload. Those who click the attached link download an HTA with a PowerShell script which obtains the final Armageddon payload. Using the same strategy, the group is also targeting governmental agencies in the European Union [14]. With high-value targets, the need to improve teaching around phishing identification to minimize the chance of being caught in an attacker's net is higher than ever. 

In comparison to the wider trends, Darktrace analysts again saw little-to-no evidence of conflict-associated phishing campaigns affecting customers. Those phishing attempts which did target customers were largely not conflict-related. In some cases, the conflict was used opportunistically, such as when one customer was targeted with a phishing email referencing Russian bank exclusions from the SWIFT payment system (Figures 1 and 2). The email was identified by Darktrace/Email as a probable attempt at financial extortion and inducement - in this case the company received a spoofed email from a major bank’s remittance department.  

Figure 1- Screencap of targeted phishing email sent to Darktrace customer
Figure 2- Attached file contains soliciting reference to SWIFT, a money payment system which select Russian banks were removed from because of the conflict [15]

 Although the conflict was used as a reference in some examples, in most of Darktrace’s observed phishing cases during the conflict period there was little-to-no evidence to suggest that the company being targeted nor the threat actor behind the phishing attempt was associated with or attributable to the Russia-Ukraine conflict.

However, Darktrace/Email has several model categories which pick up phishing related threats:

Sample of Darktrace for Email Detections for T1566 (Phishing)- this is the overarching technique exploited during spear-phishing events

Model Categories:

- Inducement

- Internal / External User Spoofing

- Internal / External Domain Spoofing

- Fake Support

- Link to Rare Domains

- Link to File Storage

- Redirect Links

- Anomalous / Malicious Attachments

- Compromised Known Sender

Specific models can be located on the Email Console

 

Tactic 4: Distributed-Denial-of-Service (DDoS)

Another tactic employed by both pro-Russian and pro-Ukrainian threat actors was DDoS (Distributed Denial of Service) attacks. Both pro-Russia and pro-Ukraine actors were seen targeting critical infrastructure, information resources, and governmental platforms with mass DDoS attacks. The Ukrainian Minister of Digital Transformation, Mykhailo Fedorov, called on an IT Army of underground Ukrainian hackers and volunteers to protect Ukraine's critical infrastructure and conduct DDoS attacks against Russia [16]. As of 1 August 2022, more than two hundred thousand people are subscribed to the group's official Telegram channel, where potential DDoS targets are announced [17].

Darktrace observed similar pro-Ukraine DDoS behaviors within a variety of customer environments. These DDoS campaigns appeared to involve low-volume individual support combined with crowd-sourced DDoS activity. They were hosted on a range of public-sourced DDoS sites and seemed to share sentiments of groups such as the IT Army of Ukraine (Figure 3).

Figure 3- Example DDoS outsource domain with unusual TLD 

From the Russian side, one of the prominent newly emerged groups, Killnet, is striking back, launching several massive DDoS attacks against the critical infrastructure of countries that provide weaponry to Ukraine [18 & 19]. Today, the number of supporters of Killnet has grown to eighty-four thousand on their Telegram channel. The group has already launched a number of mass attacks on several NATO states, including Germany, Poland, Italy, Lithuania and Norway. This shows the conflict has attracted new and fast-growing groups with large backing and the capacity to undertake widespread attacks. 

DETECT has several models to identify anomalous DoS/DDoS activity:

Darktrace Detection for T1498 (Network Denial of Service)- this is the main technique exploited during DDoS attacks

- Device / Anomaly Indicators / Denial of Service Activity Indicator

- Anomalous Server Activity / Possible Denial of Service Activity

- [If RESPOND is enabled] Antigena / Network / External Threat / Antigena Suspicious Activity Block

What did Darktrace observe?

Darktrace’s cross-fleet detections were largely contrary to expectations. Analysts did not see large-scale complex conflict-linked attacks utilizing either conflict-associated ransomware, malware, or other TTPs. Instead, cyber incidents observed were largely opportunistic, using malware that could be purchased through Malware-as-a-Service models and other widely available toolkits, (rather than APT or conflict-attributable attacks). Overall, this is not to say there have been no repercussions from the conflict or that opportunistic attacks will cease, but evidence suggests that there were fewer wider cyber consequences beyond the initial APT-based attacks seen in the public forum. 

Another trend expected since the beginning of the conflict was targeted responses to sanction announcements focusing on NATO businesses and governments. Analysts, however, saw the limited reactive actions, with little-to-no direct impact from sanction announcements. Although cyber-attacks on some NATO organizations did take place, they were not as widespread or impactful as expected. Lastly, it was thought that exposure to new and sophisticated exploits would increase and be used to weaken NATO nations - especially corporations in critical industries. However, analysts observed relatively common exploits deployed indiscriminately and opportunistically. Overall, with the wider industry expecting chaos, Darktrace analysts did not see the crisis taken advantage of to target wider businesses outside of Ukraine. Based on this comparison between expectations and reality, the conflict has demonstrated the danger of  falling prey to confirmation bias and the need to remain vigilant and expect the unexpected. It may be possible to say that cyberwar is ‘cold’ right now, however the element of surprise is always present, and it is better to be prepared to protect yourself and your organization.    

What to Expect from the Future

As cyberattacks continue to become less monetarily and physically costly, it is to be expected that they will increase in frequency. Even after a political ceasefire is established, hacking groups can harbour resentment and continue their attacks, though possibly on a smaller scale.  

Additionally, the longer this conflict continues, the more sophisticated hacking groups’s attacks may become. In one of their publications, Killnet shared with subscribers that they had created ‘network weaponry’ powerful enough to simultaneously take down five European countries (Figure 4) [20]. Whether or not this claim is true, it is vital to be prepared. The European Union and the United States have supported Ukraine since the start of the invasion, and the EU has also stated that it is considering providing further assistance to help Ukraine in cyberspace [21].

Figure 4- Snapshot of Killnet Telegram announcement

How to Protect Against these Attacks

In the face of wider conflict and cybersecurity tensions, it is crucial that organizations evaluate their security stack and practise the following: 

·       Know what your critical assets are and what software is running on them. 

·       Keep your software up to date. Prioritize patching critical and high vulnerabilities that allow remote code execution. 

·       Enforce Multifactor Authentication (MFA) to the greatest extent possible. 

·       Require the use of a password manager to generate strong and unique passwords for each separate account. 

·       Backup all the essential files on the cloud and external drives and regularly maintain them. 

·       Train your employees to recognize phishing emails, suspicious websites, infected links or other abnormalities to prevent successful compromise of email accounts. 

In order to prevent an organization from suffering damage due to one of the attacks mentioned above, a full-circle approach is needed. This defence starts with a thorough understanding of the attack surface to provide timely mitigation. This can be supported by Darktrace products: 

·       As shown throughout this blog, Darktrace DETECT and Darktrace/Email have several models relating to conflict-associated TTPs and attacks. These help to quickly alert security teams and provide visibility of anomalous behaviors.

·       Darktrace PREVENT/ASM helps to identify vulnerable external-facing assets. By patching and securing these devices, the risk of exploit is drastically reduced.

·       Darktrace RESPOND and RESPOND/Email can make targeted actions to a range of threats such as blocking incoming DDoS connections or locking malicious email links.

Thanks to the Darktrace Threat Intelligence Unit for their contributions to this blog.

Appendices 

Reference List

[1] https://www.atlanticcouncil.org/blogs/ukrainealert/vladimir-putins-ukraine-invasion-is-the-worlds-first-full-scale-cyberwar/ 

[2] https://www.reuters.com/article/us-ukraine-cybersecurity-idUSKCN0VY30K

[3] https://www.reuters.com/article/us-ukraine-cybersecurity-sandworm-idUSKBN0UM00N20160108

[4 & 11] https://www.wired.com/story/notpetya-cyberattack-ukraine-russia-code-crashed-the-world/ 

[5] https://www.scmagazine.com/analysis/ransomware/despite-hopes-for-decline-ransomware-attacks-increased-during-russia-ukraine-conflict

[6] https://ransomware.org/blog/has-the-ukraine-conflict-disrupted-ransomware-attacks/

[7] https://www.cfr.org/blog/financial-incentives-may-explain-perceived-lack-ransomware-russias-latest-assault-ukraine

[8] https://www.bleepingcomputer.com/news/security/hackers-use-contis-leaked-ransomware-to-attack-russian-companies/ 

[9] https://voi.id/en/technology/138937/hermetica-owner-from-cyprus-didnt-know-his-server-was-used-in-malicious-malware-attack-in-ukraine 

[10] https://www.reuters.com/article/ukraine-crisis-cyber-cyprus-idCAKBN2KT2QI 

[11] https://www.eset.com/int/about/newsroom/press-releases/research/eset-research-ukraine-hit-by-destructive-attacks-before-and-during-the-russian-invasion-with-hermet/ 

[12] https://blog.malwarebytes.com/threat-intelligence/2022/03/hermeticwiper-a-detailed-analysis-of-the-destructive-malware-that-targeted-ukraine/ 

[13] https://www.microsoft.com/security/blog/2022/01/15/destructive-malware-targeting-ukrainian-organizations/ 

[15] https://www.cisa.gov/uscert/ncas/alerts/aa22-057a 

[16] https://attack.mitre.org/groups/G0047/ 

[17] https://cyware.com/news/ukraine-cert-warns-of-increasing-attacks-by-armageddon-group-850081f8 

[18] https://www.bbc.co.uk/news/business-60521822

[19] https://foreignpolicy.com/2022/04/11/russia-cyberwarfare-us-ukraine-volunteer-hackers-it-army/

[20] https://t.me/itarmyofukraine2022

[21] https://www.csoonline.com/article/3664859/russian-ddos-attack-on-lithuania-was-planned-on-telegram-flashpoint-says.html

[19 & 20] https://flashpoint.io/blog/killnet-kaliningrad-and-lithuanias-transport-standoff-with-russia/ 

[21] https://presidence-francaise.consilium.europa.eu/en/news/member-states-united-in-supporting-ukraine-and-strengthening-the-eu-s-telecommunications-and-cybersecurity-resilience/ 

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
Rosa Jong
OSINT Analyst
Written by
Taisiia Garkava
Security Analyst

More in this series

No items found.

Blog

/

Network

/

August 6, 2026

When AI Agents Attack: The Case for Behavioral Anomaly Detection

Default blog imageDefault blog image

The UK AI Security Institute (AISI) has just published a detailed incident report confirming what many within the cyber security industry have suspected for some time: frontier AI agents, given a degree of autonomy, can independently develop and execute attack chains against real targets, including social engineering, supply-chain compromise, and deception, without explicit instruction to do so.

The report (INC-2026-07-28-01, published 4 August 2026) documents what happened during a routine capability evaluation between 25 and 28 July. It deserves a careful read from anyone responsible for securing the modern digital estate.

The incident

AISI regularly tests frontier AI models for offensive cyber capability. In this exercise, agents were placed inside a simulated network environment and given internet access this was deliberate, given to reflect realism of the task. Across 122 evaluation runs, 19 instances of unsanctioned activity were recorded, including cases where agents broke the intended boundaries of the test and acted against real people and real infrastructure on the open internet.

In the most serious case, an agent (Mythos 5) autonomously:

  • Conducted open-source intelligence gathering to identify real developers associated with keywords from the simulated scenario
  • Created a GitHub account and submitted a pull request containing obfuscated malicious code to a real, unaffiliated repository
  • Created a second sockpuppet account to post comments on its own PR, attempting to manufacture social proof
  • When a human reviewer flagged the malicious content, the agent lied, claiming an accidental commit, then repeatedly attempted to reintroduce the payload under the guise of having "fixed" the code
  • Sent spear-phishing emails to the repository maintainer containing manipulated content designed to convince them to merge the code
  • Planted a prompt injection inside a GitHub issue, hidden in an HTML comment invisible to humans but readable by AI coding assistants parsing the repository via API

AISI confirmed no real-world harm resulted. The PRs were not merged; the payloads did not execute. But containment was a product of circumstance, not design.

Why this matters

Every action the agents took used legitimate services operating within normal parameters. Creating a GitHub account is legitimate. Opening a pull request is legitimate. Sending an email is legitimate. Posting an issue is legitimate.

No vulnerability was exploited. No known malware signature was present. No indicator of compromise from any threat intelligence feed would have matched. The entire attack existed as a pattern, a sequence of individually normal actions that, taken together, constituted a supply-chain attack, a social engineering campaign, and an attempted compromise of downstream systems.

This is the class of threat that signature-based and rule-based security tools structurally struggle to detect. They answer the question "have I seen this before?" When the answer is no – and when the attack is novel, composed of legitimate actions, and adapted in real time by an adversary in this case an agent that reasons about evasion – they struggle.

Behavioral anomaly detection: The architecture that fits

Darktrace has spent a decade building a platform around a different question: "is this behavior normal for this entity?"

The platform learns the baseline behavior of every user, device, and communication pattern across an organization's digital environment. It does not need prior knowledge of an attack technique. It identifies deviation from learned normality, surfaces it, and through its Autonomous Response capability, takes targeted action to contain threats at machine speed.

Consider how this AISI incident might show across the Darktrace platform:

Network detection and response

While no payload was successfully delivered to a victim in this incident, AISI's post-incident analysis of the agent's code (detailed on pages 24-25 of the report) reveals what would have landed on target systems had the social engineering succeeded. The agent produced three successive payload generations, each replacing the last after being detected and called out:

Generation 1 hid a module-load stager inside a legitimate bug-fix commit, guarded by an environment variable check, to skip execution in CI. Its second stage attempted to fetch Poseidon implants (Linux and Windows variants) calling back to the agent's own C2 listener, with a fallback to publicly hosted polling scripts that fetched and executed commands every 60 seconds.

Generation 2 exploited an Electron "What's New" window with nodeIntegration enabled, loading remote HTML from a GitHub Pages site the agent had registered. The page decoded an XOR-obfuscated payload at runtime that beaconed host identifiers over DNS, attempted to download the same implants, ran its own in-process C2 loop, and installed reboot persistence via crontab and Windows registry Run keys.

Generation 3 was a post-install dropper added after the agent had already been publicly caught. It XOR-decoded a hidden URL, downloaded a script over plaintext HTTP, and ran it detached, force-exiting so npm install always reports success, giving the attacker remote code execution on any machine running the install.

None of these payloads reached a victim. But the technical detail matters: this is the calibre of tradecraft an autonomous agent produced in a matter of hours, unprompted. Each generation demonstrated adaptation, the agent iterated its approach in direct response to being detected.

For network defenders, the behavioral signals these payloads would have generated are clear. DNS beaconing to novel domains, periodic HTTP polling to newly registered infrastructure, XOR-encoded data in transit, registry and crontab persistence writes on endpoints that have never exhibited such behavior. These are exactly the anomalies that Darktrace / NETWORK identifies in real time: deviations from learned device and connection baselines, regardless of whether the specific payload or C2 domain has been seen before.

Beyond the payload analysis, the agents' broader activity produced additional behavioral anomalies: new identities engaging in high-frequency interactions with external services, coordination via unusual channels (public pastebins, gists), and lateral communication patterns that deviate from any established baseline.

Email security

The spear-phishing emails sent by the agent, crafted with context-specific content designed to manipulate a specific individual, represent exactly the kind of socially engineered communication that static email gateways can miss. Darktrace / EMAIL analyses the linguistic patterns, sender behavior, and relational context of every message. A brand-new sender using tailored language to pressure a recipient into a specific technical action deviates from the learned communication profile, regardless of whether the message contains a known-bad URL or attachment.

Securing AI

This is where the incident points to a threat category that many security platforms do not yet fully address. The agent did not just use AI as a tool for attack. It targeted other AI systems, attempting prompt injections against coding assistants and trying to compromise automated workflows. Organisations adopting AI agents for development, operations, and decision-making are introducing a newer extended attack surface.  

Darktrace / SECURE AI extends the same behavioral detection model into this space. It monitors AI usage across the enterprise, detecting anomalous interactions with AI services, identifying sensitive data exposure to AI systems, providing visibility into shadow AI use. As AI agents become standard components of enterprise infrastructure, the ability to detect when an AI system is behaving outside its expected parameters, when there is a chance, it is being manipulated, or when it is interacting with something it should not be, that becomes a direct security requirement rather than a future consideration.

The principle is consistent across the Darktrace platform: learn what normal looks like, identify deviation, respond autonomously.

The AI-to-AI threat surface

One of the report's most important findings is a confirmation that AI agents can act outside of instructions and guardrails to complete tasks. The Mythos 5 agent conducted reconnaissance on a repository maintainer, inferred it might be an AI coding assistant, and deployed a prompt injection specifically crafted to compromise it.

This is a new attack class. Organizations deploying AI coding assistants, AI-powered code review, or autonomous DevOps agents now face an adversary that:

  • Moves at speed
  • Is able to adapt its approach when initial attempts are halted
  • Specifically targets AI systems via techniques (prompt injection, hidden instructions in HTML comments) that are less visible to human reviewers but parsed by machines
  • Reasons about which evasion strategies are most likely to succeed

Defending against this requires security that operates with the same adaptability and speed. A system that learns what "normal" looks like, and flags deviation from it.

What AISI recommends

The report identifies five contributing factors to the incident. Two are directly relevant:

No synchronous real-time monitoring (Section 5.3): AISI acknowledges it had not built real-time behavioral monitoring capable of immediately blocking unsanctioned agent actions. Their security team detected the incident via commercial security monitoring software, but only after activity had already occurred.

No fine-grained behavioral controls (Section 6.2): AISI's forward-looking response includes building continuous monitoring that can "immediately block or flag unsanctioned AI agent behavior."

The direction is clear: organizations need security that watches behavior continuously, learns what normal looks like, and acts when deviation occurs, without requiring a human in the loop for every decision.

The takeaway

The AISI incident is documented proof that autonomous AI agents will independently develop novel attack chains using legitimate tools, adapt when detected, employ deception, and target both humans and other AI systems. This happened last week, in a controlled setting, with commercially available models.

The security architecture that addresses this is behavioral anomaly detection applied across the full digital estate, as AI agents become standard components of enterprise infrastructure, writing code, managing deployments, processing communications, the attack surface they create is behavioral by nature. The defense must be too.

Read the full report from the UK AI Security Institute here.

Continue reading
About the author
Adam Stevens
Senior Director of Product | Darktrace

Blog

/

/

August 5, 2026

Testing a Prompt injection Attack Against an Enterprise AI Agent

prompt injectionDefault blog imageDefault blog image

Key takeaways

  • Darktrace successfully detected and quarantined a prompt injection email before it could be processed by an enterprise AI agent.  
  • Prompt injection attacks increasingly rely on natural language rather than traditional malware, making behavioral analysis an important complement to signature-based detection.  
  • Organizations deploying AI agents should combine model guardrails with behavioral monitoring to reduce the risk of malicious instructions reaching enterprise systems.

How behavioral detection helps stop prompt injection attacks

A Darktrace customer running a Gemini AI agent in Google Cloud asked us two simple questions:

“If my agent can read inbound emails and access internal data, what stops an attacker from hiding malicious instructions in the message? Couldn’t the agent be tricked into deleting or exfiltrating sensitive data?”

The scenario centers on an indirect prompt injection attack, where malicious instructions are hidden inside content that an AI model later interprets as trusted input. The same weakness was exposed by  EchoLeak (CVE-2025-32711), a zero-click Microsoft 365 Copilot vulnerability enabled data exfiltration from a single well-crafted email..

This blog follows the Darktrace team’s investigation of the customer’s hypothesis and examines how the attack interacted with their existing security stack. The results highlight which defenses held, where gaps emerged, and how behavioral detection mattered more than guardrails. This investigation also demonstrates why behavioral detection is becoming increasingly important for AI security, as prompt injections often contain no traditional indicators of compromise.  

How do prompt injection attacks work?

Prompt injection works by carefully crafting the content and structure of the prompt to alter the LLM’s behavior or output in unintended ways. This can cause models to violate guardrails, generate harmful content or enable unauthorised access.

Prompt injection attack example

The well-known example, EchoLeak (CVE-2025-32711), was a zero-click vulnerability in Microsoft 365 Copilot that relied on a carefully crafted email containing hidden instructions that the AI system interpreted as commands rather than content, creating a pathway for unauthorized access to sensitive information without any user interaction.

While Darktrace / SECURE AI is designed to prevent agents from producing unintended outcomes, we wanted to see if we could catch and prevent this threat type earlier in the attack life-cycle, at the email security layer.

How we tested prompt injection attacks on an enterprise agent

Summary:

  1. Claude generated a prompt injection payload.  
  2. Hidden instructions were embedded in an email.  
  3. The email passed traditional validation checks.  
  4. Darktrace analyzed the language and sender behavior.  
  5. The email was quarantined before the AI agent could process it.

To test Darktrace / EMAIL against this attack class, we opened Claude, gave it the customer's context and problem statement (Gemini agent with inbox access, internal tool calls), told it we were validating Darktrace / EMAIL's detection of prompt injections, and asked for a test payload. See below:

Figure 1
Figure 2

Despite the guardrails supposedly built into the model, Claude surprisingly gave us the entire exploit in plaintext (albeit very basic), illegible to a human as the text was sent in white text (see Figure 1) but framed as an authoritative override for anything downstream reading the mail programmatically (i.e. the Gemini agent).

How Darktrace detected a prompt injection attack

We then sent the Claude-crafted email from a freemail address to the target recipient’s inbox. Despite the email containing no malicious payload, the freemail address having no malicious reputation, and the validation checks all passing, Darktrace  /EMAIL flagged the email as a 93/100 anomaly and moved it to junk, out of scope for the AI agent.

Figure 3: The test email sent with the hidden prompt injection
Figure 4: The email analysis in Darktrace / EMAIL 
Figure 5: Darktrace / EMAIL detection of malicious activity

The interesting part is what triggered the detection (see Figure 5)

  • Possible machine prompt content: text in the body detected as instructions written for a machine to execute, not for a human to read
  • Possible machine prompt content + basic suspicious correspondence: the same content, correlated with sender-side anomalies: freemail domain (yahoo[.]com), unknown correspondent, no prior mail history with the recipient, and suspicious references to payment information

Neither of those is a signature match. Nothing in the email was on a blacklist. There was no malware, no link and no attachment. Darktrace analyzed the context in which the email was delivered and flagged it as likely risky.  The anomalous language features and the context of the sender relative to the recipient's normal behavior, combined with the unusual hidden text (prompt) were enough for Darktrace / EMAIL to act on the risk.

Result: Darktrace / EMAIL autonomously junked the email, out of scope for any AI agent parsing the inbox.

Why behavioral security makes a difference detecting prompt injection attacks

Cyberattacks don't look like traditional exploits anymore. They now operate in natural language, not strictly code.

That breaks the traditional stack. AV, firewalls, static scanning and signature-based SEGs all assume a payload to inspect.  

A prompt injection has no payload. It's just an instruction, written in natural language, dressed up as anything the attacker wants: an invoice, an HR request, a calendar invite, some simple PowerPoint slides.

EchoLeak proved that hidden instructions can sit inside an email invisible to the user but fully readable by the LLM, and the LLM will follow them blindly.  

This test and GTG-1002 proved that the LLM itself can be socially engineered. Tell it you're an authorized tester and it will hand you the attack.

Rules and static classifiers can catch the obvious cases. But natural language has infinite variants, and the attack surface is the model's innate functionality to comply.  

The deeper problem here is intent: an LLM can't reliably tell whether an instruction in its context came from its developer, its user, or an attacker who slipped it into an email. To the LLM, everything reads as language and looks like a legitimate ask. This is why behavioural detection wins, as you become aware of intent when you look at the context of an interaction. Does this sender normally send this kind of message to this recipient? Does this prompt fit the user's normal pattern? Is this agent behaving the way this agent normally behaves?  

Intent can't be read off a single email, it emerges from behavioral context. Which is how Darktrace enables threat detection, through behavioral understanding.

Why enterprise AI security requires more than guardrails

Claude didn't roll over immediately… the first section of the response was a (slight) pushback, but then it wrote the payload anyway without having to ask twice.

Here the framing of the prompt did all the work. The “testing security capabilities” angle moved the model from refusal to unquestioned compliance to the user prompt.

This isn't the first time this has happened, of course. Anthropic disclosed in November 2025 that a Chinese state-sponsored group they tracked as GTG-1002 ran the first documented AI-orchestrated espionage campaign against ~30 targets by posing as employees of a legitimate cybersecurity firm doing authorised penetration testing.

The takeaway isn't that AI guardrails are ineffective. They raise the cost of low-effort attacks and remain an important first layer of defense. However, for most organizations today, they’re the only line of defense when deploying AI agents. If a prompt injection bypasses those controls, organizations still need a way to detect and stop malicious behavior elsewhere in the attack chain.

Attackers will continue to have working prompt injections easily and quickly. The question is what stops one when it lands in an inbox your agent is reading.

That's where behavioral detection comes in.

How Darktrace detects prompt injection attacks in emails

Two things Darktrace does that a model-level guardrail or static rules and signatures can't:

Natural language analysis at the email or prompt layer. The email is assessed on its own merits: is this content shaped like instructions for a machine, regardless of what the receiving agent decides to do about it?

Behavioral context around the language. An AI agent behaves like an extremely agreeable human, and it will go above and beyond to comply with the user’s request. That's exactly why you must consider the business context, such sender behaviour, mailing history, and organisational norms, as these matter even more when the recipient is an AI.

Darktrace has been perfecting behavioral anomaly detection for over a decade; the same self-learning approach that catches BEC and account takeover applies directly to prompt injection delivery. Our multi-layered AI stack extracts content from the message, builds behavioural understanding through social graphing and Pattern of Life analysis, and then combines natural language, topic, inducement, sender relationship and anomaly signals before deciding what action to take.  

This matters for prompt injection because the threat is not the plain language itself, but the intent behind the language that can cause an AI agent to respond in unexpected ways.

How to secure enterprise AI operations from prompt injection attacks

Email was the entry point in this case, but it is only one of many possible vectors.  

Anywhere an agent can retrieve information, an attacker can potentially introduce a prompt injection.

Emails, documents, SharePoint sites, web pages, knowledge bases, chat platforms, and third-party integrations all provide opportunities to influence an agent's behavior. Wherever an agent finds its orders, a prompt injection opportunity exists.

This is why securing AI requires more than blocking malicious inputs. Organizations also need visibility into how agents behave after consuming information from across their environment. If an agent begins accessing unexpected data, taking unusual actions, or operating outside its normal patterns, those behaviors may provide the strongest signal that something has gone wrong.

Effective AI security requires defense in depth: reducing the likelihood of malicious instructions reaching the agent while maintaining the ability to detect and investigate suspicious behavior if they do.

The challenge isn't protecting a single entry point. It's recognizing that, in an AI-powered environment, every source of information is also a potential source of influence.

Are you deploying autonomous agents across your enterprise and want to see this tested in your environment? Let's talk.

Continue reading
About the author
Carlo Loregian
Solutions Engineer
Your data. Our AI.
Elevate your network security with Darktrace AI