Blog
/
Network
/
September 5, 2022

Steps of a BumbleBee Intrusion to a Cobalt Strike

Discover the steps of a Bumblebee intrusion, from initial detection to Cobalt Strike deployment. Learn how Darktrace defends against evolving threats with AI.
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
Default blog image
05
Sep 2022

Introduction

Throughout April 2022, Darktrace observed several cases in which threat actors used the loader known as ‘BumbleBee’ to install Cobalt Strike Beacon onto victim systems. The threat actors then leveraged Cobalt Strike Beacon to conduct network reconnaissance, obtain account password data, and write malicious payloads across the network. In this article, we will provide details of the actions threat actors took during their intrusions, as well as details of the network-based behaviours which served as evidence of the actors’ activities.  

BumbleBee 

In March 2022, Google’s Threat Analysis Group (TAG) provided details of the activities of an Initial Access Broker (IAB) group dubbed ‘Exotic Lily’ [1]. Before March 2022, Google’s TAG observed Exotic Lily leveraging sophisticated impersonation techniques to trick employees of targeted organisations into downloading ISO disc image files from legitimate file storage services such as WeTransfer. These ISO files contained a Windows shortcut LNK file and a BazarLoader Dynamic Link Library (i.e, DLL). BazarLoader is a member of the Bazar family — a family of malware (including both BazarLoader and BazarBackdoor) with strong ties to the Trickbot malware, the Anchor malware family, and Conti ransomware. BazarLoader, which is typically distributed via email campaigns or via fraudulent call campaigns, has been known to drop Cobalt Strike as a precursor to Conti ransomware deployment [2]. 

In March 2022, Google’s TAG observed Exotic Lily leveraging file storage services to distribute an ISO file containing a DLL which, when executed, caused the victim machine to make HTTP requests with the user-agent string ‘bumblebee’. Google’s TAG consequently called this DLL payload ‘BumbleBee’. Since Google’s discovery of BumbleBee back in March, several threat research teams have reported BumbleBee samples dropping Cobalt Strike [1]/[3]/[4]/[5]. It has also been reported by Proofpoint [3] that other threat actors such as TA578 and TA579 transitioned to BumbleBee in March 2022.  

Interestingly, BazarLoader’s replacement with BumbleBee seems to coincide with the leaking of the Conti ransomware gang’s Jabber chat logs at the end of February 2022. On February 25th, 2022, the Conti gang published a blog post announcing their full support for the Russian state’s invasion of Ukraine [6]. 

Figure 1: The Conti gang's public declaration of their support for Russia's invasion of Ukraine

Within days of sharing their support for Russia, logs from a server hosting the group’s Jabber communications began to be leaked on Twitter by @ContiLeaks [7]. The leaked logs included records of conversations among nearly 500 threat actors between Jan 2020 and March 2022 [8]. The Jabber logs were supposedly stolen and leaked by a Ukrainian security researcher [3]/[6].

Affiliates of the Conti ransomware group were known to use BazarLoader to deliver Conti ransomware [9]. BumbleBee has now also been linked to the Conti ransomware group by several threat research teams [1]/[10]/[11]. The fact that threat actors’ transition from BazarLoader to BumbleBee coincides with the leak of Conti’s Jabber chat logs may indicate that the transition occurred as a result of the leaks [3]. Since the transition, BumbleBee has become a significant tool in the cyber-crime ecosystem, with links to several ransomware operations such as Conti, Quantum, and Mountlocker [11]. The rising use of BumbleBee by threat actors, and particularly ransomware actors, makes the early detection of BumbleBee key to identifying the preparatory stages of ransomware attacks.  

Intrusion Kill Chain 

In April 2022, Darktrace observed the following pattern of threat actor activity within the networks of several Darktrace clients: 

1.     Threat actor socially engineers user via email into running a BumbleBee payload on their device

2.     BumbleBee establishes HTTPS communication with a BumbleBee C2 server

3.     Threat actor instructs BumbleBee to download and execute Cobalt Strike Beacon

4.     Cobalt Strike Beacon establishes HTTPS communication with a Cobalt Strike C2 server

5.     Threat actor instructs Cobalt Strike Beacon to scan for open ports and to enumerate network shares

6.     Threat actor instructs Cobalt Strike Beacon to use the DCSync technique to obtain password account data from an internal domain controller

7.     Threat actor instructs Cobalt Strike Beacon to distribute malicious payloads to other internal systems 

With limited visibility over affected clients’ email environments, Darktrace was unable to determine how the threat actors interacted with users to initiate the BumbleBee infection. However, based on open-source reporting on BumbleBee [3]/[4]/[10]/[11]/[12]/[13]/[14]/[15]/[16]/[17], it is likely that the actors tricked target users into running BumbleBee by sending them emails containing either a malicious zipped ISO file or a link to a file storage service hosting the malicious zipped ISO file. These ISO files typically contain a LNK file and a BumbleBee DLL payload. The properties of these LNK files are set in such a way that opening them causes the corresponding DLL payload to run. 

In several cases observed by Darktrace, devices contacted a file storage service such as Microsoft OneDrive or Google Cloud Storage immediately before they displayed signs of BumbleBee infection. In these cases, it is likely that BumbleBee was executed on the users’ devices as a result of the users interacting with an ISO file which they were tricked into downloading from a file storage service. 

Figure 2: The above figure, taken from the event log for an infected device, shows that the device contacted a OneDrive endpoint immediately before making HTTPS connections to the BumbleBee C2 server, 45.140.146[.]244
Figure 3: The above figure, taken from the event log for an infected device, shows that the device contacted a Google Cloud Storage endpoint and then the malicious endpoint ‘marebust[.]com’ before making HTTPS connections to the  BumbleBee C2 servers, 108.62.118[.]61 and 23.227.198[.]217

After users ran a BumbleBee payload, their devices immediately initiated communications with BumbleBee C2 servers. The BumbleBee samples used HTTPS for their C2 communication, and all presented a common JA3 client fingerprint, ‘0c9457ab6f0d6a14fc8a3d1d149547fb’. All analysed samples excluded domain names in their ‘client hello’ messages to the C2 servers, which is unusual for legitimate HTTPS communication. External SSL connections which do not specify a destination domain name and whose JA3 client fingerprint is ‘0c9457ab6f0d6a14fc8a3d1d149547fb’ are potential indicators of BumbleBee infection. 

Figure 4:The above figure, taken from Darktrace's Advanced Search interface, depicts an infected device's spike in HTTPS connections with the JA3 client fingerprint ‘0c9457ab6f0d6a14fc8a3d1d149547fb’

Once the threat actors had established HTTPS communication with the BumbleBee-infected systems, they instructed BumbleBee to download and execute Cobalt Strike Beacon. This behaviour resulted in the infected systems making HTTPS connections to Cobalt Strike C2 servers. The Cobalt Strike Beacon samples all had the same JA3 client fingerprint ‘a0e9f5d64349fb13191bc781f81f42e1’ — a fingerprint associated with previously seen Cobalt Strike samples [18]. The domain names ‘fuvataren[.]com’ and ‘cuhirito[.]com’ were observed in the samples’ HTTPS communications. 

Figure 5:The above figure, taken from Darktrace's Advanced Search interface, depicts the Cobalt Strike C2 communications which immediately followed a device's BumbleBee C2 activity

Cobalt Strike Beacon payloads call home to C2 servers for instructions. In the cases observed, threat actors first instructed the Beacon payloads to perform reconnaissance tasks, such as SMB port scanning and SMB enumeration. It is likely that the threat actors performed these steps to inform the next stages of their operations.  The SMB enumeration activity was evidenced by the infected devices making NetrShareEnum and NetrShareGetInfo requests to the srvsvc RPC interface on internal systems.

Figure 6: The above figure, taken from Darktrace’s Advanced Search interface, depicts a spike in srvsvc requests coinciding with the infected device's Cobalt Strike C2 activity

After providing Cobalt Strike Beacon with reconnaissance tasks, the threat actors set out to obtain account password data in preparation for the lateral movement phase of their operation. To obtain account password data, the actors instructed Cobalt Strike Beacon to use the DCSync technique to replicate account password data from an internal domain controller. This activity was evidenced by the infected devices making DRSGetNCChanges requests to the drsuapi RPC interface on internal domain controllers. 

Figure 7: The above figure, taken from Darktrace’s Advanced Search interface, depicts a spike in DRSGetNCChanges requests coinciding with the infected device’s Cobalt Strike C2 activity

After leveraging the DCSync technique, the threat actors sought to broaden their presence within the targeted networks.  To achieve this, they instructed Cobalt Strike Beacon to get several specially selected internal systems to run a suspiciously named DLL (‘f.dll’). Cobalt Strike first established SMB sessions with target systems using compromised account credentials. During these sessions, Cobalt Strike uploaded the malicious DLL to a hidden network share. To execute the DLL, Cobalt Strike abused the Windows Service Control Manager (SCM) to remotely control and manipulate running services on the targeted internal hosts. Cobalt Strike first opened a binding handle to the svcctl interface on the targeted destination systems. It then went on to make an OpenSCManagerW request, a CreateServiceA request, and a StartServiceA request to the svcctl interface on the targeted hosts: 

·      Bind request – opens a binding handle to the relevant RPC interface (in this case, the svcctl interface) on the destination device

·      OpenSCManagerW request – establishes a connection to the Service Control Manager (SCM) on the destination device and opens a specified SCM database

·      CreateServiceA request – creates a service object and adds it to the specified SCM database 

·      StartServiceA request – starts a specified service

Figure 8: The above figure, taken from Darktrace’s Advanced Search interface, outlines an infected system’s lateral movement activities. After writing a file named ‘f.dll’ to the C$ share on an internal server, the infected device made several RPC requests to the svcctl interface on the targeted server

It is likely that the DLL file which the threat actors distributed was a Cobalt Strike payload. In one case, however, the threat actor was also seen distributing and executing a payload named ‘procdump64.exe’. This may suggest that the threat actor was seeking to use ProcDump to obtain authentication material stored in the process memory of the Local Security Authority Subsystem Service (LSASS). Given that ProcDump is a legitimate Windows Sysinternals tool primarily used for diagnostics and troubleshooting, it is likely that threat actors leveraged it in order to evade detection. 

In all the cases which Darktrace observed, threat actors’ attempts to conduct follow-up activities after moving laterally were thwarted with the help of Darktrace’s SOC team. It is likely that the threat actors responsible for the reported activities were seeking to deploy ransomware within the targeted networks. The steps which the threat actors took to make progress towards achieving this objective resulted in highly unusual patterns of network traffic. Darktrace’s detection of these unusual network activities allowed security teams to prevent these threat actors from achieving their disruptive objectives. 

Darktrace Coverage

Once threat actors succeeded in tricking users into running BumbleBee on their devices, Darktrace’s Self-Learning AI immediately detected the command-and-control (C2) activity generated by the loader. BumbleBee’s C2 activity caused the following Darktrace models to breach:

·      Anomalous Connection / Anomalous SSL without SNI to New External

·      Anomalous Connection / Suspicious Self-Signed SSL

·      Anomalous Connection / Rare External SSL Self-Signed

·      Compromise / Suspicious TLS Beaconing To Rare External

·      Compromise / Beacon to Young Endpoint

·      Compromise / Beaconing Activity To External Rare

·      Compromise / Sustained SSL or HTTP Increase

·      Compromise / Suspicious TLS Beaconing To Rare External

·      Compromise / SSL Beaconing to Rare Destination

·      Compromise / Large Number of Suspicious Successful Connections

·      Device / Multiple C2 Model Breaches 

BumbleBee’s delivery of Cobalt Strike Beacon onto target systems resulted in those systems communicating with Cobalt Strike C2 servers. Cobalt Strike Beacon’s C2 communications resulted in breaches of the following models: 

·      Compromise / Beaconing Activity To External Rare

·      Compromise / High Volume of Connections with Beacon Score

·      Compromise / Large Number of Suspicious Successful Connections

·      Compromise / Sustained SSL or HTTP Increase

·      Compromise / SSL or HTTP Beacon

·      Compromise / Slow Beaconing Activity To External Rare

·      Compromise / SSL Beaconing to Rare Destination 

The threat actors’ subsequent port scanning and SMB enumeration activities caused the following models to breach:

·      Device / Network Scan

·      Anomalous Connection / SMB Enumeration

·      Device / Possible SMB/NTLM Reconnaissance

·      Device / Suspicious Network Scan Activity  

The threat actors’ attempts to obtain account password data from domain controllers using the DCSync technique resulted in breaches of the following models: 

·      Compromise / Unusual SMB Session and DRS

·      Anomalous Connection / Anomalous DRSGetNCChanges Operation

Finally, the threat actors’ attempts to internally distribute and execute payloads resulted in breaches of the following models:

·      Compliance / SMB Drive Write

·      Device / Lateral Movement and C2 Activity

·      Device / SMB Lateral Movement

·      Device / Multiple Lateral Movement Model Breaches

·      Anomalous File / Internal / Unusual SMB Script Write

·      Anomalous File / Internal / Unusual Internal EXE File Transfer

·      Anomalous Connection / High Volume of New or Uncommon Service Control

If Darktrace/Network had been configured in the targeted environments, then it would have blocked BumbleBee’s C2 communications, which would have likely prevented the threat actors from delivering Cobalt Strike Beacon into the target networks. 

Figure 9: Attack timeline

Conclusion

Threat actors use loaders to smuggle more harmful payloads into target networks. Prior to March 2022, it was common to see threat actors using the BazarLoader loader to transfer their payloads into target environments. However, since the public disclosure of the Conti gang’s Jabber chat logs at the end of February, the cybersecurity world has witnessed a shift in tradecraft. Threat actors have seemingly transitioned from using BazarLoader to using a novel loader known as ‘BumbleBee’. Since BumbleBee first made an appearance in March 2022, a growing number of threat actors, in particular ransomware actors, have been observed using it.

It is likely that this trend will continue, which makes the detection of BumbleBee activity vital for the prevention of ransomware deployment within organisations’ networks. During April, Darktrace’s SOC team observed a particular pattern of threat actor activity involving the BumbleBee loader. After tricking users into running BumbleBee on their devices, threat actors were seen instructing BumbleBee to drop Cobalt Strike Beacon. Threat actors then leveraged Cobalt Strike Beacon to conduct network reconnaissance, obtain account password data from internal domain controllers, and distribute malicious payloads internally.  Darktrace’s detection of these activities prevented the threat actors from achieving their likely harmful objectives.  

Thanks to Ross Ellis for his contributions to this blog.

Appendices 

References 

[1] https://blog.google/threat-analysis-group/exposing-initial-access-broker-ties-conti/ 

[2] https://securityintelligence.com/posts/trickbot-gang-doubles-down-enterprise-infection/ 

[3] https://www.proofpoint.com/us/blog/threat-insight/bumblebee-is-still-transforming

[4] https://www.cynet.com/orion-threat-alert-flight-of-the-bumblebee/ 

[5] https://research.nccgroup.com/2022/04/29/adventures-in-the-land-of-bumblebee-a-new-malicious-loader/ 

[6] https://www.bleepingcomputer.com/news/security/conti-ransomwares-internal-chats-leaked-after-siding-with-russia/ 

[7] https://therecord.media/conti-leaks-the-panama-papers-of-ransomware/ 

[8] https://www.secureworks.com/blog/gold-ulrick-leaks-reveal-organizational-structure-and-relationships 

[9] https://www.prodaft.com/m/reports/Conti_TLPWHITE_v1.6_WVcSEtc.pdf 

[10] https://www.kroll.com/en/insights/publications/cyber/bumblebee-loader-linked-conti-used-in-quantum-locker-attacks 

[11] https://symantec-enterprise-blogs.security.com/blogs/threat-intelligence/bumblebee-loader-cybercrime 

[12] https://isc.sans.edu/diary/TA578+using+thread-hijacked+emails+to+push+ISO+files+for+Bumblebee+malware/28636 

[13] https://isc.sans.edu/diary/rss/28664 

[14] https://www.logpoint.com/wp-content/uploads/2022/05/buzz-of-the-bumblebee-a-new-malicious-loader-threat-report-no-3.pdf 

[15] https://ghoulsec.medium.com/mal-series-23-malware-loader-bumblebee-6ab3cf69d601 

[16]  https://blog.cyble.com/2022/06/07/bumblebee-loader-on-the-rise/  

[17]  https://asec.ahnlab.com/en/35460/ 

[18] https://thedfirreport.com/2021/07/19/icedid-and-cobalt-strike-vs-antivirus/

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

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