Blog
/
Network
/
November 13, 2024

Tactics Behind the Royal and Blacksuit Ransomware

Delve into the complexities of the Royal and Blacksuit ransomware strains and their implications for cybersecurity in today’s digital landscape.
Written by
Signe Zaharka
Principal Cyber Analyst
Inside the SOC
Darktrace cyber analysts are world-class experts in threat intelligence, threat hunting and incident response, and provide 24/7 SOC support to thousands of Darktrace customers around the globe. Inside the SOC is exclusively authored by these experts, providing analysis of cyber incidents and threat trends, based on real-world experience in the field.
Written by
Signe Zaharka
Principal Cyber Analyst
Default blog image
13
Nov 2024

What is BlackSuit Ransomware?

Since late 2023, Darktrace has detected BlackSuit ransomware infiltrating multiple customer networks in the US. This ransomware has targeted a wide range of industries, including arts, entertainment, real estate, public administration, defense, and social security.

Emerging in May 2023, BlackSuit is believed to be a spinoff of Royal ransomware due to similarities in code and Conti, and most likely consists of Russian and Eastern European hackers [1]. Recorded Future reported that the ransomware had affected 95 organizations worldwide, though the actual number is likely much higher [2]. While BlackSuit does not appear to focus on any particular sector, it has targeted multiple organizations in the healthcare, education, IT, government, retail and manufacturing industries [3]. Employing double extortion tactics, BlackSuit not only encrypts files but also steals sensitive data to leverage ransom payments.

BlackSuit has demanded over USD 500 million in ransoms, with the highest individual demand reaching USD 60 million [4]. Notable targets include CDK Global, Japanese media conglomerate Kadokawa, multiple educational institutions, Octapharma Plasma, and the government of Brazil [5][6][7][8].

Darktrace’s Coverage of BlackSuit Ransomware Attack

Case 1, November 2023

The earliest attack on a Darktrace customer by BlackSuit was detected at the start of November 2023. The unusual network activity began on a weekend—a time commonly chosen by ransomware groups to increase their chances of success, as many security teams operate with reduced staff. Darktrace identified indicators of the attackers’ presence on the network for almost two weeks, during which a total of 15 devices exhibited suspicious behavior.

The attack commenced with unusual internal SMB (Server Message Block) connections using a compromised service account. An internal device uploaded an executable (zzza.exe) to a domain controller (DC) and shortly after, wrote a script (socks5.ps1) to another device. According to a Cybersecurity Advisory from the CISA (Cybersecurity and Infrastructure Security Agency, US), the script file was a PowerShell reverse proxy [9].

Approximately an hour and a half later, the device to which the script was written exhibited uncommon WMI (Windows Management Instrumentation) activity. Two hours after receiving the executable file, the DC was observed making an outgoing NTLM request, using PowerShell to remotely execute commands, distributing differently named executable files (<PART OF THE CUSTOMER’S NAME>.exe), and controlling services on other devices.

Eighteen hours after the start of the unusual activity, Darktrace detected another device making repeated connections to “mystuff.bublup[.]com”, which the aforementioned CISA Advisory identifies as a domain used by BlackSuit for data exfiltration [9].

About ten minutes after the suspicious executables were distributed across the network, and less than 24 hours after the start of the unusual activity, file encryption began. A total of ten devices were seen appending the “.blacksuit” extension to files saved on other devices using SMB, as well as writing ransom notes (readme.blacksuit.txt). The file encryption lasted less than 20 minutes.

An example of the contents of a BlackSuit ransom note being written over SMB.
Figure 1: An example of the contents of a BlackSuit ransom note being written over SMB.

During this compromise, external connections to endpoints related to ConnectWise’s ScreenConnect remote management tool were also seen from multiple servers, suggesting that the tool was likely being abused for command-and-control (C2) activity. Darktrace identified anomalous connectivity associated with ScreenConnect was seen up to 11 days after the start of the attack.

10 days after the start of the compromise, an account belonging to a manager was detected adding “.blacksuit” extensions to the customer’s Software-a-Service (SaaS) resources while connecting from 173.251.109[.]106. Six minutes after file encryption began, Darktrace flagged the unusual activity and recommended a block. However, since Autonomous Response mode was not enabled, the customer’s security team needed to manually confirm the action. Consequently, suspicious activity continued for about a week after the initial encryption. This included disabling authentication on the account and an unusual Teams session initiated from the suspicious external endpoint 216.151.180[.]147.

Case 2, February 2024

Another BlackSuit compromise occurred at the start of February 2024, when Darktrace identified approximately 50 devices exhibiting ransomware-related activity in another US customer’s environment. Further investigation revealed that a significant number of additional devices had also been compromised. These devices were outside Darktrace’s purview to the customer’s specific deployment configuration. The threat actors managed to exfiltrate around 4 TB of data.

Initial access to the network was gained via a virtual private network (VPN) compromise in January 2024, when suspicious connections from a Romanian IP address were detected. According to CISA, the BlackSuit group often utilizes the services of initial access brokers (IAB)—actors who specialize in infiltrating networks, such as through VPNs, and then selling that unauthorized access to other threat actors [9]. Other initial access vectors include phishing emails, RDP (Remote Desktop Protocol) compromise, and exploitation of vulnerable public-facing applications.

Similar to the first case, the file encryption began at the end of the working week. During this phase of the attack, affected devices were observed encrypting files on other internal devices using two compromised administrator accounts. The encryption activity lasted for approximately six and a half hours. Multiple alerts were sent to the customer from Darktrace’s Security Operations Centre (SOC) team, who began reviewing the activity within four minutes of the start of the file encryption.

Darktrace’s Cyber AI Analyst clustering together multiple events related to unusual activity on the network, including file encryption over SMB by BlackSuit.
Figure 2: Darktrace’s Cyber AI Analyst clustering together multiple events related to unusual activity on the network, including file encryption over SMB by BlackSuit.
Figure 3: A spike in model alerts on the day when file encryption by BlackSuit was observed in the network.

In this case, the threat actor utilized SystemBC proxy malware for command and control (C2). A domain controller (DC) was seen connecting to 137.220.61[.]94 on the same day the file encryption took place. The DC was also observed connecting to a ProxyScrape domain around the same time, which is related to the SOCKS5 protocol used by SystemBC. During this compromise, RDP, SSH, and SMB were used for lateral movement within the network.

Figure 4: A Cyber AI Analyst investigation alerting to a device on the VPN subnet making suspicious internal SSH connections due to malicious actors moving laterally within the network.

Signs of threat actors potentially being on the network were observed as early as two days prior to the file encryption. This included unusual internal network scanning via multiple protocols (ICMP, SMB, RDP, etc.), credential brute-forcing, SMB access failures, and anonymous SMBv1 sessions. These activities were traced to IP addresses belonging to two desktop devices in the VPN subnet associated with two regular employee user accounts. Threat actors were seemingly able to exploit at least one of these accounts due to LDAP legacy policies being in place on the customer’s environment.

A Cyber AI Analyst incident summary alerting to a device on the VPN subnet conducting internal reconnaissance.
Figure 5: A Cyber AI Analyst incident summary alerting to a device on the VPN subnet conducting internal reconnaissance.
Examples of the proposed Darktrace Autonomous Response actions on the day BlackSuit initiated file encryption.
Figure 6: Examples of the proposed Darktrace Autonomous Response actions on the day BlackSuit initiated file encryption.

Case 3, August 2024

The most recently observed BlackSuit compromise occurred in August 2024, when a device was observed attempting to brute-force the credentials of an IT administrator. This activity continued for 11 days.

Once the admin’s account was successfully compromised, network scanning, unusual WMI, and SAMR (Security Account Manager Remote protocol) activity followed. A spike in the use of this account was detected on a Sunday—once again, the attackers seemingly targeting the weekend—when the account was used by nearly 50 different devices.

The compromised admin’s account was exploited for data gathering via SMB, resulting in the movement of 200 GB of data between internal devices in preparation for exfiltration. The files were then archived using the naming convention “*.part<number>.rar”.

Around the same time, Darktrace observed data transfers from 19 internal devices to “bublup-media-production.s3.amazonaws[.]com,” totaling just over 200 GB—the same volume of data gathered internally. Connections to other Bublup domains were also detected. The internal data download and external data transfer activity took approximately 8-9 hours.

Unfortunately, Darktrace was not configured in Autonomous Response mode at the time of the attack, meaning any mitigative actions to stop the data gathering or exfiltration required human confirmation.  

One of the compromised devices was seen sending 80 GB of data to bublup-media-production.s3.amazonaws[.]com within a span of 4 hours.
Figure 7: One of the compromised devices was seen sending 80 GB of data to bublup-media-production.s3.amazonaws[.]com within a span of 4 hours.

Once the information was stolen, the threat actor moved on to the final stage of the attack—file encryption. Five internal devices, using either the compromised admin account or connecting via anonymous SMBv1 sessions, were seen encrypting files and writing ransom notes to five other devices on the network. The attempts at file encryption continued for around two hours, but Darktrace’s Autonomous Response capability was able to block the activity and prevent the attack from escalating.

Conclusion

The persistent and evolving threat posed by ransomware like BlackSuit underscores the critical importance of robust cybersecurity measures across all sectors. Since its emergence in 2023, BlackSuit has demonstrated a sophisticated approach to infiltrating networks, leveraging double extortion tactics, and demanding substantial ransoms. The cases highlighted above illustrate the varied methods and persistence of BlackSuit attackers, from exploiting VPN vulnerabilities to abusing remote management tools and targeting off-hours to maximize impact.

Although many similar connection patterns, such as the abuse of Bublup services for data exfiltration or the use of SOCKS5 proxies for C2, were observed during cases investigated by Darktrace, BlackSuit actors are highly sophisticated and tailors their attacks to each target organization. The consequences of a successful attack can be highly disruptive, and remediation efforts can be time-consuming and costly. This includes taking the entire network offline while responding to the incident, restoring encrypted files from backups (if available), dealing with damage to the organization’s reputation, and potential lawsuits.

These BlackSuit ransomware incidents emphasize the need for continuous vigilance, timely updates to security protocols, and the adoption of autonomous response technologies to swiftly counteract such attacks. As ransomware tactics continue to evolve, organizations must remain agile and informed to protect their critical assets and data. By learning from these incidents and enhancing their cybersecurity frameworks, organizations can better defend against the relentless threat of ransomware and ensure the resilience of their operations in an increasingly digital world.

Credit to Signe Zaharka (Principal Cyber Analyst) and Adam Potter (Senior Cyber Analyst)

Insights from Darktrace’s First 6: Half-year threat report for 2024

First 6: half year threat report darktrace screenshot

Darktrace’s First 6: Half-Year Threat Report 2024 highlights the latest attack trends and key threats observed by the Darktrace Threat Research team in the first six months of 2024.

  • Focuses on anomaly detection and behavioral analysis to identify threats
  • Maps mitigated cases to known, publicly attributed threats for deeper context
  • Offers guidance on improving security posture to defend against persistent threats

‍

‍

Appendices

Darktrace Model Detections

Anomalous Connection / Data Sent to Rare Domain

Anomalous Connection / High Volume of New or Uncommon Service Control

Anomalous Connection / New or Uncommon Service Control

Anomalous Connection / Rare WinRM Outgoing

Anomalous Connection / SMB Enumeration

Anomalous Connection / Suspicious Activity On High Risk Device

Anomalous Connection / Suspicious Read Write Ratio

Anomalous Connection / Suspicious Read Write Ratio and Unusual SMB

Anomalous Connection / Sustained MIME Type Conversion

Anomalous Connection / Uncommon 1 GiB Outbound

Anomalous Connection / Unusual Admin SMB Session

Anomalous File / Internal / Additional Extension Appended to SMB File

Anomalous File / Internal / Executable Uploaded to DC

Anomalous File / Internal / Unusual SMB Script Write

Anomalous Server Activity / Anomalous External Activity from Critical Network Device

Anomalous Server Activity / Outgoing from Server

Anomalous Server Activity / Rare External from Server

Anomalous Server Activity / Write to Network Accessible WebRoot

Compliance / Outgoing NTLM Request from DC

Compliance / Remote Management Tool On Server

Compliance / SMB Drive Write

Compromise / Beacon to Young Endpoint

Compromise / Beaconing Activity To External Rare

Compromise / Ransomware / Possible Ransom Note Read

Compromise / Ransomware / Possible Ransom Note Write

Compromise / Ransomware / SMB Reads then Writes with Additional Extensions

Compromise / Ransomware / Suspicious SMB Activity

Device / Anomalous RDP Followed By Multiple Model Breaches

Device / EXE Files Distributed to Multiple Devices

Device / Internet Facing Device with High Priority Alert

Device / Large Number of Model Breaches

Device / Large Number of Model Breaches from Critical Network Device

Device / Multiple Lateral Movement Model Breaches

Device / Network Scan

Device / New or Uncommon WMI Activity

Device / New or Unusual Remote Command Execution

Device / New User Agent To Internal Server

Device / SMB Lateral Movement

Device / SMB Session Brute Force (Admin)

Device / Suspicious SMB Scanning Activity

Device / Unusual LDAP Query For Domain Admins

SaaS / Access / Teams Activity from Rare Endpoint

SaaS / Resource / SaaS Resources With Additional Extensions

SaaS / Unusual Activity / Disabled Strong Authentication

SaaS / Unusual Activity / Multiple Unusual SaaS Activity Scores

SaaS / Unusual Activity / Unusual SaaS Activity Score

SaaS / Unusual Activity / Unusual Volume of SaaS Modifications

Unusual Activity / Anomalous SMB Delete Volume

Unusual Activity / Anomalous SMB Move & Write

Unusual Activity / High Volume Client Data Transfer

Unusual Activity / High Volume Server Data Transfer

Unusual Activity / Internal Data Transfer

Unusual Activity / SMB Access Failures

Unusual Activity / Sustained Anomalous SMB Activity

Unusual Activity / Unusual External Data to New Endpoint

User / New Admin Credentials on Client

User / New Admin Credentials on Server

User/ Kerberos Password Bruteforce

Autonomous Response Models

Antigena / Network / External Threat / Antigena File then New Outbound Block

Antigena / Network / External Threat / Antigena Ransomware Block

Antigena / Network / External Threat / Antigena Suspicious Activity Block

Antigena / Network / External Threat / SMB Ratio Antigena Block

Antigena / Network / Insider Threat / Antigena Internal Anomalous File Activity

Antigena / Network / Insider Threat / Antigena Internal Data Transfer Block

Antigena / Network / Insider Threat / Antigena Large Data Volume Outbound Block

Antigena / Network / Insider Threat / Antigena Network Scan Block

Antigena / Network / Insider Threat / Antigena Unusual Privileged User Activities Block

Antigena / Network / Insider Threat / Antigena Unusual Privileged User Activities Pattern of Life Block

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

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

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

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

Antigena / Network / Significant Anomaly / Antigena Significant Anomaly from Client Block

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

Antigena / Network / Significant Anomaly / Repeated Antigena Breaches

Antigena / SaaS / Antigena Unusual Activity Block

List of Indicators of Compromise (IoCs)

IoC - Type - Description + Confidence

.blacksuit - File extension – When encrypting the files, this extension is appended to the filename – High

readme.blacksuit.txt – ransom note - A file demanding cryptocurrency payment in exchange for decrypting the victim's files and not leaking the stolen data – High

mystuff.bublup[.]com, bublup-media-production.s3.amazonaws[.]com – data exfiltration domains related to an organization and project management app that has document sharing functionality – High

137.220.61[.]94:4001 – SystemBC C2 related IP address (this tool is often used by other ransomware groups as well) - Medium

173.251.109[.]106 – IP address seen during a SaaS BlackSuit compromise (during file encryption) – Medium

216.151.180[.]147 – IP address seen during a SaaS BlackSuit compromise (during an unusual Teams session) - Medium

MITRE ATT&CK Mapping

Tactic - Technqiue

Account Manipulation - PERSISTENCE - T1098

Alarm Suppression - INHIBIT RESPONSE FUNCTION - T0878

Application Layer Protocol - COMMAND AND CONTROL - T1071

Automated Collection - COLLECTION - T1119

Block Command Message - INHIBIT RESPONSE FUNCTION - T0803

Block Reporting Message - INHIBIT RESPONSE FUNCTION - T0804

Browser Extensions - PERSISTENCE - T1176

Brute Force I/O - IMPAIR PROCESS CONTROL - T0806

Brute Force - CREDENTIAL ACCESS - T1110

Client Configurations - RECONNAISSANCE - T1592.004 - T1592

Cloud Accounts - DEFENSE EVASION, PERSISTENCE, PRIVILEGE ESCALATION, INITIAL ACCESS - T1078.004 - T1078

Data Destruction - IMPACT - T1485

Data Destruction - INHIBIT RESPONSE FUNCTION - T0809

Data Encrypted for Impact - IMPACT - T1486

Data from Cloud Storage Object - COLLECTION - T1530

Data Staged - COLLECTION - T1074

Domain Groups - DISCOVERY - T1069.002 - T1069

Email Collection - COLLECTION - T1114

Exfiltration Over C2 Channel - EXFILTRATION - T1041

Exfiltration to Cloud Storage - EXFILTRATION - T1567.002 - T1567

Exploit Public - Facing Application - INITIAL ACCESS - T1190

Exploitation for Privilege Escalation - PRIVILEGE ESCALATION - T0890

Exploitation of Remote Services - LATERAL MOVEMENT - T1210

File and Directory Discovery - DISCOVERY - T1083

File Deletion - DEFENSE EVASION - T1070.004 - T1070

IP Addresses - RECONNAISSANCE - T1590.005 - T1590

Lateral Tool Transfer - LATERAL MOVEMENT - T1570

LLMNR/NBT - NS Poisoning and SMB Relay - CREDENTIAL ACCESS, COLLECTION - T1557.001 - T1557

Modify Alarm Settings - INHIBIT RESPONSE FUNCTION - T0838

Modify Control Logic - IMPAIR PROCESS CONTROL, INHIBIT RESPONSE FUNCTION - T0833

Modify Parameter - IMPAIR PROCESS CONTROL - T0836

Network Service Scanning - DISCOVERY - T1046

Network Share Discovery - DISCOVERY - T1135

Pass the Hash - DEFENSE EVASION, LATERAL MOVEMENT - T1550.002 - T1550

RDP Hijacking - LATERAL MOVEMENT - T1563.002 - T1563

Remote Access Software - COMMAND AND CONTROL - T1219

Remote Desktop Protocol - LATERAL MOVEMENT - T1021.001 - T1021

Remote System Discovery - DISCOVERY - T1018

Rename System Utilities - DEFENSE EVASION - T1036.003 - T1036

Scanning IP Blocks - RECONNAISSANCE - T1595.001 - T1595

Scheduled Transfer - EXFILTRATION - T1029

Service Execution - EXECUTION - T1569.002 - T1569

Service Stop - IMPACT - T1489

SMB/Windows Admin Shares - LATERAL MOVEMENT - T1021.002 - T1021

Stored Data Manipulation - IMPACT - T1565.001 - T1565

Taint Shared Content - LATERAL MOVEMENT - T1080

Valid Accounts - DEFENSE EVASION, PERSISTENCE, PRIVILEGE ESCALATION, INITIAL ACCESS - T1078

Vulnerability Scanning - RECONNAISSANCE - T1595.002 - T1595

Web Protocols - COMMAND AND CONTROL - T1071.001 - T1071

Web Services - RESOURCE DEVELOPMENT - T1583.006 - T1583

Web Shell - PERSISTENCE - T1505.003 - T1505

Windows Management Instrumentation - EXECUTION - T1047

Windows Remote Management - LATERAL MOVEMENT - T1021.006 - T1021

References

1.     https://www.trendmicro.com/en_us/research/23/e/investigating-blacksuit-ransomwares-similarities-to-royal.html

2.     https://www.reuters.com/technology/cybersecurity/blacksuit-hacker-behind-cdk-global-attack-hitting-us-car-dealers-2024-06-27/

3.     https://www.sentinelone.com/anthology/blacksuit/

4.     https://thehackernews.com/2024/08/fbi-and-cisa-warn-of-blacksuit.html

5.     https://www.techtarget.com/whatis/feature/The-CDK-Global-outage-Explaining-how-it-happened

6.     https://therecord.media/japanese-media-kadokawa-investigating-cyber

7.     https://therecord.media/plasma-donation-company-cyberattack-blacksuit

8.     https://thecyberexpress.com/government-of-brazil-cyberattack-by-blacksuit/

9.     https://www.cisa.gov/news-events/cybersecurity-advisories/aa23-061a

Written by
Signe Zaharka
Principal Cyber Analyst
Inside the SOC
Darktrace cyber analysts are world-class experts in threat intelligence, threat hunting and incident response, and provide 24/7 SOC support to thousands of Darktrace customers around the globe. Inside the SOC is exclusively authored by these experts, providing analysis of cyber incidents and threat trends, based on real-world experience in the field.
Written by
Signe Zaharka
Principal Cyber Analyst

Botnet Behind the Camera: Mirai Katana Activity on a Video Recording Device

OT
•
September 2, 2026
Parvatha Ananthakannan
Cyber Analyst
Watch the NIS2 Webinar
Trending blogs
1
Prompt Security in Enterprise AI: Strengths, Weaknesses, and Common Approaches
May 20, 2026
2
How Empowering End Users can Improve Your Email Security and Decrease the Burden on the SOC
May 8, 2024
3
Elevating Network Security: Confronting Trust, Ransomware, & Novel Attacks
Jun 21, 2024
4
The State of AI in Cybersecurity: Unveiling Global Insights from 1,800 Security Practitioners
Apr 9, 2024
5
The State of AI in Cybersecurity: The Impact of AI on Cybersecurity Solutions
May 13, 2024

More in this series

No items found.
Continue reading
Network
•
August 6, 2026

When AI Agents Attack: The Case for Behavioral Anomaly Detection

A new UK AI Security Institute report shows autonomous AI agents independently developing novel attack chains, adapting when detected, and targeting humans and other AI systems. As agents become embedded across the enterprise, behavioral anomaly detection will be critical to securing the new attack surface they create.
Adam Stevens
Senior Director of Product
Read more
Network
•
July 27, 2026

Hiding in Plain Sight: Uncovering a Multi-Stage Ransomware Attack Through Behavioral Detection

Darktrace detected a multi-stage ransomware intrusion from its earliest stages, identifying reconnaissance, privilege escalation, lateral movement, command-and-control communications, and data exfiltration before encryption occurred. This analysis highlights how behavioral detection exposed attacker activity using legitimate tools and infrastructure, providing multiple opportunities for early intervention and containment.
Alexandra Evzona
Cyber Analyst
Read more
Network
•
July 1, 2026

9 Stages of Ransomware & How AI Responds

Discover the 9 stages of ransomware attacks and how AI responds at each stage. Learn how you can protect your business from cyber threats.
The Darktrace Community
Read more

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