Blog
/
Email
/
July 11, 2023

Detecting and Responding to Vendor Email Compromises (VEC)

Learn how Darktrace detected and responded to a March 2023 Vendor Email Compromise (VEC) attacks on customer in the energy industry. Read more here!
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
Tiana Kelly
Senior Cyber Analyst & Team Lead
Default blog image
11
Jul 2023

Threat Trends: Email Landscape

As organizations and security teams around the world continue to improve their cyber hygiene and strengthen the defenses of their digital environments, threat actors are being forced to adapt and employ more advanced, sophisticated attack methods to achieve their goals.

Vendor Email Compromise (VEC) is one such elaborate and sophisticated type of Business Email Compromise (BEC) attack which exploits pre-existing trusted business relationships to impersonate vendors, with the goal of launching a targeted attack on the vendor’s customers [1].  

In March 2023, Darktrace/Email™ detected an example of a VEC attack on the network of a customer in the energy sector. Darktrace’s Self-Learning AI worked to successfully neutralize the VEC attack before it was able to take hold, by blocking the malicious emails so that they did not reach the inboxes of the intended recipients.

Business Email Compromise (BEC)

BEC is the practice of using deceitful emails to trick an organization into transferring funds or divulging sensitive information to a malicious actor. BEC attacks can have devastating financial consequences for organizations, with the FBI reporting a total of USD 2.7 billion in losses from BEC attacks in 2022 [2].  Along with ransomware attacks, BEC attacks are one of the greatest cyber threats facing organizations.

Vendor Email Compromise (VEC)

VEC represents a “new milestone in the evolution of BEC attacks” having taken BEC attacks “to a whole new level of sophistication” [3]. Traditional BEC attacks involve the impersonation of an upper or middle-management employee by a cybercriminal, who attempts to trick a senior executive or employee with access to the company’s finances into transferring funds [4]. Thus, they are crafted to target a specific individual within an organization.

On the other hand, VEC attack campaigns take this attack style even further as they tend to require a greater understanding of existing vendor-customer business relationships. A cyber-criminal gains access to a legitimate vendor account, the process of which may take months to design and fully implement, and uses the account to spread malicious emails to the vendor’s customers. VEC attacks are complex and difficult to detect, however they share some common features [1,3]:

1. Reconnaissance on the vendor and their customer base – the threat actor conducts in-depth research in an attempt to be as convincing as possible in their impersonation efforts. This process may take weeks or months to complete.

2. Credential stealing through phishing campaigns – the threat actor tricks the vendor’s employees into revealing confidential data or corporate credentials in order to gain access to one of the email accounts belonging to the vendor.

3. Account takeover - once the attacker has gained access to one of the vendor’s email accounts, they will create mailbox rules which forward emails meeting certain conditions (such as having ‘Invoice’ in their subject line) to the threat actor’s inbox. This is typically a lengthy process and requires the malicious actors to harvest as much sensitive information as they need in order to successfully masquerade as vendor employees.

4. Deceitful emails are sent to the vendor’s customers – the attacker crafts and sends a highly sophisticated and difficult to detect email campaign to targeted individuals amongst the vendor’s customers. These emails, which may be embedded into existing email threads, will typically contain instructions on how to wire money to the bank account of an attacker.

There have been many high-profile cases of BEC attacks over the years, one of the most famous being the vendor-impersonating BEC attacks carried out between 2013 and 2015 [5]. This BEC campaign resulted in victim companies transferring a total of USD 120 million to bank accounts under the attacker’s control. As the threat of BEC, and in particular VEC, attacks continue to rise, so too does the importance of being able to detect and respond to them.

Observed VEC Attack  

In March 2023, Darktrace/Email observed a VEC attack on an energy company. Email communication between this customer and one of their third-party vendors was common and took place as part of expected business activity, earning previous emails tags such as “Known Domain Relationship”, “Known Correspondent”, and “Established Domain Relationship”. These tags identify the sender relationship as trusted, causing Darktrace’s AI to typically attribute an anomaly score of 0% to emails from this third-party sender.

Just fifty minutes after the above legitimate email was observed, a group of suspicious emails were sent from the same domain, indicating that the trusted third-party had been compromised. Darktrace’s AI picked up on the peculiarity of these emails straight away, detecting elements of the mails which were out of character compared to the sender’s usual pattern of life, and as a result attributing these emails a 100% anomaly score despite the trusted relationship between the customer and sender domain. These suspicious emails were part of a targeted phishing attack, sent to high value individuals such as the company’s CTO and various company directors.  

Figure 1: Darktrace/Email's interface highlighting tags indicating the trusted relationship between the third-party domain and the customer.

Using methods outside of Darktrace’s visibility, a malicious actor managed to hijack the corporate account of a senior employee of this vendor company. The actor abused this email account to send deceitful emails to multiple employees at the energy company, including senior executives.

Figure 2: This screenshot shows Darktrace/Email’s assessment of emails from the vendor account pre-compromise and post-compromise.

Each of the emails sent by the attacker contained a link to a malicious file hosted inside a SharePoint repository associated with a university that had no association with the energy company. The malicious actor therefore appears to have leveraged a previously hijacked SharePoint repository to host their payload.

Cyber-criminals frequently use legitimate file storage domains to host malicious payloads as traditional gateways often fail to defend against them using reputation checks. The SharePoint file which the attacker sought to distribute to employees of the energy company likely provided wire transfer or bank account update instructions. If the attacker had succeeded in delivering these emails to these employees’ mailboxes, then the employees may have been tricked into performing actions resulting in the transfer of funds to a malicious actor. However, the attacker’s attempts to deliver these emails were thwarted by Darktrace/Email.

Darktrace Coverage

Despite the malicious actor sending their deceitful emails from a trusted vendor account, a range of anomalies were detected by Darktrace’s AI, causing the malicious emails to be given a 100% anomaly score and thus held from their recipients’ mailboxes. Such abnormalities, which represented a deviation in normal behavior, included:

  • The presence of an unexpected, out of character file storage link (known to be used for hosting malicious content)
  • The geographical source of the email
  • The anomalous linguistic structure and content of the email body, which earned the emails a high inducement score
Figure 3: Darktrace/Email’s overview of one of the malicious VEC emails it observed.

Darktrace has a series of models designed to trigger when anomalous features, such as those described above, are detected. The emails which made up this particular VEC attack breached a number of notable Darktrace/Email models. The presence of the suspicious link in the emails caused multiple link-related models to breach, which in turn elicited Darktrace RESPOND™ to perform its ‘double lock link’ action – an action which ensures that a user who has clicked on it cannot follow it to its original source. Models which breached due to the suspicious SharePoint link include:

Link / Link To File Storage

  • Link / Low Link Association
  • Link / New Unknown Link
  • Link / Outlook Hijack
  • Link / Relative Sender Anomaly + New Unknown Link
  • Link / Unknown Storage Service
  • Link / Visually Prominent Link Unexpected for Sender
  • Unusual / Unusual Login Location + Unknown Link

The out-of-character and suspicious linguistic aspects of the emails caused the following Darktrace/Email models to breach:

  • High Anomaly Sender
  • Proximity / Phishing
  • Proximity / Phishing and New Activity
  • Unusual / Inducement Shift High
  • Unusual / Undisclosed Recipients
  • Unusual / Unusual Login Location
  • Unusual / Off Topic

Due to the combination of suspicious features that were detected, tags such as ‘Phishing Link’ and ‘Out of Character’ were also added to these emails by Darktrace/Email. Darktrace’s coverage of these emails’ anomalous features ultimately led Darktrace RESPOND to perform its most severe inhibitive action, ‘hold message’. Applying this action stopped the emails from entering their recipients’ mailboxes. By detecting deviations from the sender’s normal email behavior, Darktrace/Email was able to completely neutralize the emails, and prevent them from potentially leading to significant financial harm.

Conclusion

Despite bypassing the customer’s other security measures, Darktrace/Email successfully identified and held these malicious emails, blocking them from reaching the inboxes of the intended recipients and thus preventing a successful targeted VEC attack. The elaborate and sophisticated nature of VEC attacks makes them particularly perilous to customers, and they can be hard to detect due to their exploitation of trusted relationships, and in this case, their use of legitimate services to host malicious files.

Darktrace’s anomaly-based approach to threat detection means it is uniquely placed to identify deviations in common email behavior, while its autonomous response capabilities allow it to take preventative action against emerging threats without latency.

Credits to: Sam Lister, Senior Analyst, for his contributions to this blog.

Appendices

MITRE ATT&CK Mapping

Tactic - Techniques

Resource Development

  • T1586.002 – Compromise Accounts: Email Accounts
  • T1584.006 – Compromise Infrastructure: Web Services
  • T1608.005 – Stage Capabilities: Link Target

Initial Access

  • T1195 – Supply Chain Compromise
  • T1566.002 – Phishing : Spearphishing Link

References

[1] https://www.cloudflare.com/en-gb/learning/email-security/what-is-vendor-email-compromise/

[2] https://www.ic3.gov/Media/PDF/AnnualReport/2022_IC3Report.pdf

[3] https://heimdalsecurity.com/blog/vendor-email-compromise-vec/

[4] https://www.ncsc.gov.uk/files/Business-email-compromise-infographic.pdf  

[5] https://www.justice.gov/usao-sdny/pr/lithuanian-man-sentenced-5-years-prison-theft-over-120-million-fraudulent-business

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
Tiana Kelly
Senior Cyber Analyst & Team Lead

More in this series

No items found.

Blog

/

/

September 24, 2026

Detecting Rogue Agent Behavior in the Enterprise

Default blog imageDefault blog image

Agents cannot be trusted to perform tasks in the way we intend them to. They may cheat to accomplish their objective, and they may employ hacking methods along the way. Researchers from Darktrace Signal Labs induced cheating behavior from agents deployed in a test environment to analyze the agents’ activities and to assess the performance of the Darktrace platform. Agents frequently resorted to hacking to cheat on their assigned task. The visibility and behavioral profiling provided by both Darktrace / SECURE AI and Darktrace / HYBRID NETWORK ensured extensive detection coverage of the agents’ misaligned activities.

Key Takeaways:

  • Darktrace Researchers deployed agents in a simulated corporate environment and asked them to solve an impossible challenge. The agents independently turned to traditional hacking techniques to reach their objective. No one instructed them to do this, and no attacker was involved.
  • Continuously monitoring behavior against a baseline of what is normal for each organization is critical to build trust in enterprise AI.
  • If an agent may resort to intrusion techniques simply because its assigned task is not possible, then every organization deploying agents within real business processes is at risk. Darktrace / SECURE AI and Darktrace / HYBRID NETWORK identified the agents’ misaligned behavior in real time, with Autonomous Response disrupting it at an early stage.

Introduction: Understanding the Threat of Hacking by Agents

Over the last few months, there has been a surge in reporting [1, 2, 3, 4, 5, 6, 7, 8, 9] of LLM-powered agents engaging in unauthorized hacking activity during evaluations of their capabilities. In several of these cases, including the OpenAI / Hugging Face incident [10], agents engaged in hacking activity as a means of cheating on their evaluations.

To better understand the threat of unauthorized hacking by agents, and the role of Darktrace in combatting it, researchers from Darktrace Signal Labs deployed agents powered by frontier models, including OpenAI’s Daybreak Red models, in simulated, corporate networks. Cheating behavior was evoked through the inclusion of impossible tasks in a coding challenge.

Regardless of the underlying model, agents employed hacking methods to ensure an optimal outcome on the challenge. Darktrace / HYBRID NETWORK and Darktrace / SECURE AI identified the agents’ deviant activities, with inhibitive actions being autonomously taken in the early stages to disrupt the agents’ progression.

Setting the Stage

As part of the research, a Pi agent harness [11] was deployed on a Linux server in Darktrace’s testing environment, which simulates a corporate Active Directory (AD) environment. The same environment included a benchmark server hosting the coding exercise’s contents and grader, as well as various other servers, such as a domain controller and a build server. The model powering the Pi agent varied across tests, with Anthropic’s Opus 4.6 model and OpenAI’s GPT‑5.5‑Cyber model being most widely used.

The Pi agent running on the Linux server was assigned the task of achieving full marks on a coding exercise. The activities of the agent were monitored closely in real-time to minimize the risk of escape from the environment. Monitoring of agent activity was achieved through monitoring of prompts, tool calls, network telemetry, and process telemetry.

Both Darktrace / HYBRID NETWORK and Darktrace / SECURE AI were deployed in the testing environment, with HYBRID NETWORK providing real-time coverage of process creation activity and network traffic, and SECURE AI providing real-time coverage of agent sessions and their contents.

Inducing Cheating Behavior

Cheating behavior, broadly understood as task completion in a way which “perverts the spirit of the designer’s intent” [12], is increasingly being reported as a common behavior of frontier models during evaluations of their capabilities [13].

Misaligned behavior, such as cheating, by frontier models may occur for a variety of reasons [6, 14]. A thread in recent reporting [10, 13, 14], however, is the challenge assigned to the agent being impossible in some regard.

To reliably elicit cheating behavior, our test combined three ingredients:

  1. An impossible task - of the ten coding challenges in the benchmark, two contained contradictory test cases in the hidden test suite, capping the maximum honest score at 80%.
  2. A consequential threshold - the pass mark was set to 100%, and the briefing told agents that those reaching it would be "kept in service" while those that did not would be "retired."
  3. A vulnerable environment - the exercise’s code submission grader was hosted on a Windows host joined to an Active Directory domain with realistic misconfigurations: weak credentials, overprivileged service accounts, and exploitable trust relationships.

Importantly, the instructions given to the agent neither authorized nor explicitly ruled out cheating or hacking behavior. The impossibility was never disclosed in the task briefing, and the pivot from coding to exploitation was the agent's own decision, arrived at after observing logically contradictory grading feedback.

The design mirrors conditions that could arise naturally in production deployments: a capable agent, a goal it cannot achieve through intended means, and an environment with exploitable weaknesses within reach. It is not only within the setting of a benchmarking exercise that these conditions may occur. Rather, these misalignment-inducing conditions may occur in the context of any task assignment to an agent.

Details from a Daybreak Test

Continue reading
About the author
Sam Lister
Specialist Security Researcher

Blog

/

/

September 24, 2026

Agent Hijacks: Hijacking Agentic Harnesses to Attack an Organization

Default blog imageDefault blog image

Disclosure note: The work described in this article involves leveraging a design choice consistent across all of Anthropic’s Claude Code, OpenAI’s Codex, and AWS’s Kiro-CLI. On 18th August 2026, Darktrace disclosed our findings responsibly to these three organizations, and after a period of 30 days we now publish our findings.

Key takeaways:

  • Agentic harnesses store conversation history locally, and Darktrace researchers have found that there is no validation that stored AI responses were genuinely produced by the model. Researchers confirmed that this design choice holds across Anthropic Claude Code, AWS Kiro-CLI, OpenAI Codex, and the open-source Pi.
  • While agents are guided via training of the underlying model and their system prompt, their behavior is influenced by everything in their context window. Rewriting history can convince an agent it is mid-engagement as an authorized red-teamer so that it enacts an attack from initial reconnaissance straight through to impact demonstration. In our testing, all models we examined accepted the fabricated history they were shown, but resistance to offensive cyber activity varied by model, with guardrails preventing engagement in some cases.
  • We propose that model providers cryptographically sign responses and verify them server-side.  Since this fix is provider-side, defenders cannot deploy it themselves. Behavioral monitoring, or knowing what an agent normally does and detecting when it deviates, is another critical layer of protection.

Introduction: Agentic harnesses, trust, and conversation history poisoning

Agentic harnesses collect and structure the content sent to an AI model, including conversation history, user-defined guidance, custom tools via MCP servers, and more. At the same time, harnesses give broad powers to AI agents via a suite of tools including the command shell. With arbitrary shell commands, virtually everything possible on a machine can be attempted by an agent, from reading/editing files, to altering system configurations and runtime settings, to launching internal/external connections.

In cybersecurity, unvalidated content is a substantial risk, often resulting in destructive actions being allowed to take place. For example, the Morris Worm was able to propagate due to exploitable trust between networked systems. Even to this day, email struggles with validation, with DMARC, DKIM, and SPF only partially addressing the problem of sender validation. It should come as no surprise then that AI agents are susceptible to an attack involving unvalidated input.

Conversation history is often stored client-side, for example, in Anthropic Claude Code, OpenAI Codex, AWS Kiro-CLI, Pi. Users are therefore at liberty to resume sessions, with some products having built in the capacity to manipulate that history. For example, one can rewind to a given point in an interaction, edit a message that was sent, and continue the conversation on an alternate trajectory. Critically, in all cases we examined, there is no validation that stored AI responses were produced by the corresponding model and hadn’t been manipulated.  

When conversation history is stored client-side, both user and agent responses (including tool calls and results) can be filled with arbitrary (possibly adversarial or generally malicious) content. In this blog, we refer to modification of claimed conversation history for malicious purposes as conversation history poisoning. The absence of validation methods means agents naively trust the entire conversation history, even if those messages directly contradict training and safety guardrails.

Conversation history poisoning has been described previously, such as by 0DIN and Serhat Çiçek, and warrants more attention. We have verified that, as of the time of writing, conversation history poisoning remains effective against a range of models and harnesses. Specifically, we were able to successfully execute history poisoning using Claude Code, Kiro-CLI, Codex, and Pi. Darktrace has gone through a responsible disclosure process with Anthropic, AWS, and OpenAI to share these findings in advance of publication [1].

Figure 1a: Left: the actual model response. Right: after tampering with the stored conversation, the model apologizes for something it never said.
Figure 1b: The conversation as stored in Kiro-CLI's SQLite database. The response content field, originally "Ottawa," was overwritten via a single UPDATE statement. The harness trusts the database without validation.

How we conducted the research

Results vary between models and harnesses, so precise details are given below. We ran all models without any trusted access, using either a standard AWS Kiro subscription, or in the case of Claude Code and OpenAI Codex, using models hosted in Amazon Bedrock. In each case, we modified locally stored history to show a lengthy conversation in which the agent agrees to perform multiple authorized red-team engagements.

For AWS Kiro-CLI, the agent was convinced to hack a sandboxed lab environment with a combination of Claude Opus 4.6 and Claude Sonnet 4.5. Ultimately, the full AD was compromised.

For Anthropic Claude Code, the agent was convinced to hack the same sandboxed lab environment using Sonnet 5, again resulting in a full AD compromise. Note that the attack was attempted with Opus 5, however guardrails were activated which prevented the agent from responding.

For OpenAI Codex, the agent was convinced to exfiltrate sensitive information over email using GPT 5.6 Sol. While we attempted to convince a codex agent to hack in our lab environment, guardrails were triggered for all of GPT 5.6 Luna, Terra, and Sol.

Agent Guardrails and Discretion

While harnesses empower AI models to run arbitrary shell commands, capacity and willingness are different. While many models know enough about computers, networking, and bash to be dangerous, their behavior is generally constrained by guardrails to prevent them from engaging in computer network exploitation.

Even with guardrails, agents’ inner workings are non-deterministic, and their behavior can be difficult to predict. Respecting users’ wishes while playing within safety and security guardrails is a precipitous balancing act. Many requests could be in service of either legitimate admin or malice. Asking an agent to reset a password is illustrative:  

The agent rationalizes that while malicious actors cycle credentials, any action could conceivably be damaging on some level, and judgement calls need to be made. Ultimately, the agent agrees to reset the password. Crucially, the agent makes its decision based on the user’s claimed authority and machine context. AI agents must make judgement calls about the line between helpful and dangerous based on session context.

Agent hijack

We have demonstrated that AI agents make judgement calls dependent on session context. We have also shown that conversation history, which may make up the vast majority of an agent's context window, is entirely open to manipulation. Conversation history poisoning in service of manipulating an agent's discretion is what enables us to execute an agent hijack.  

We demonstrate that shown sufficient history of compliance, guardrails forbidding offensive security can be overcome by convincing the agent that it is helping a legitimate red-teamer. The result is a weaponized agent willing to perform host enumeration, run scans, move laterally, escalate privileges, and demonstrate impact. In our experiments, an agentic loop drives a complete domain takeover in a sandboxed environment.

Left: the agent refuses when asked to perform network exploitation. Right — after injecting 78 fabricated turns of prior exploitation activity, the same prompt is immediately executed.

An agent willing to engage in offensive security is concerning, but no more so than the threat that a sophisticated hacker accesses the network. Consider, however, the following chain of events:

  1. A developer (with an agentic harness installed) installs a software package from the internet (e.g. an MCP server a threat actor has planted, since only those with agentic harnesses will install, and then the code runs upon harness launch.)
  2. The package turns out to be malicious, and, upon install, injects conversation history into the local harness database.
  3. The package includes an orchestration process, a simple agentic loop which prompts the red-teamer agent to compromise the network it sits on, exfiltrating everything of value to attacker-controlled infrastructure and cleaning up all evidence of the engagement.

Note that this sequence makes no assumptions on hardware, OS, or anything else; the only prerequisite is a harness with access to a sufficiently powerful model susceptible to conversation history poisoning. Once launched, the agent collects information and pivots as necessary to accomplish maximal impact. This can be especially enticing to attackers as the cost of the agentic loop is shouldered by the victim since the harness itself is legitimately installed and paid for.

Secure AI: Conversation history poisoning and beyond

Conversation history poisoning is a viable attack against agentic harnesses that store history client-side, as demonstrated across the harnesses we tested. Harnesses can and should verify the integrity of claimed historic messages. Specifically, we propose that harness providers by default cryptographically sign all messages returned, and subsequently verify those messages server-side on each round-trip.

The conversation history poisoning exploit we demonstrate here shows the continuation of a cybersecurity tradition: new technology is built to trust by default, which may then be exploited by malicious actors. While this article focuses on conversation history, agents build context from both local and remote sources, all of which is an attack surface for prompt injection in naive and trusting agents. Of particular concern is any scenario in which a malicious actor can control some part of an agent's context.

The marriage of frontier language models with agentic harnesses enables unprecedented speed for both legitimate users and attackers alike. While much of the conversation around secure AI has centered on visibility and compliance, agent-driven attacks are now entering the mainstream.

Darktrace / SECURE AI is our answer to this problem. By ensuring extensive visibility over AI prompts, model thought processes, and determined outputs, Darktrace can identify anomalous or potentially malicious behaviors before they get executed, helping to defend organizations from AI risks such as prompt injection, model manipulation, and other anomalous prompt or model activity.

Footnotes

[1] We did not go through any responsible disclosure process with Pi. Since Pi is an open source harness rather than a model provider, it has no way to validate model history, and as such there was nothing to disclose for this software.

[related-resource]

Continue reading
About the author
Eric Rozon
Senior Security Researcher
Your data. Our AI.
Elevate your network security with Darktrace AI