3AM Ransomware Affiliate Breaches UK Schools in Suspected Hybrid AI Intrusion

Red | Attack
3AM Ransomware Affiliate Breaches UK Schools in Suspected Hybrid AI Intrusion | Hive Pro Threat Advisory
Threat Advisory · Attack Report

3AM Ransomware Affiliate Breaches UK Schools in Suspected Hybrid AI Intrusion

A 3AM (ThreeAM) ransomware affiliate breached a UK school estate, stole credentials from backup servers and domain controllers, built two-way Active Directory trusts and encrypted about 37 machines. Command content strongly indicates language-model authorship.

Attack ReportTA2026325Admiralty Code A13AM RansomwareThreeAMDouble ExtortionWindows
Date of publication
November 06, 2026
Admiralty code
A1
TA number
TA2026325
First seen
February 2023
Targeted platform
Windows
Malware
3AM Ransomware
Targeted regions
United States, United Kingdom, Australia, Argentina, Canada, France, Spain, Brazil, Mexico, Germany, Vietnam, Singapore, Netherlands, Luxembourg, Croatia, Malaysia, Finland, Colombia, Italy, Lebanon
Targeted industries
Manufacturing, Healthcare, Technology, Business Services & Consulting, Legal, Education, Transportation, Real Estate, Agriculture, Financial Services, Government, Aerospace and Defense, Retail, Food Service, Energy, Casino & Gambling, Associations, Media, Pharmaceutical, Hospitality

Summary

3AM, also tracked as ThreeAM, is a Rust-based ransomware-as-a-service operation active since February 2023 that reportedly began as a fallback payload for failed LockBit deployments and runs a double-extortion model, encrypting files with the .threeamtime extension, wiping shadow copies and leaking stolen data.

In its latest activity, an affiliate breached a UK school estate through an unknown access vector, stole credentials from backup servers and domain controllers, and created backdoor domain administrators. The operator then built two-way trusts across separate Active Directory domains, exfiltrated finance data over an internal C2 channel and encrypted about 37 machines. Command content strongly indicates language-model authorship, and retry timing is consistent with a human operator directing an AI agentic loop.


Attack Details

#1

3AM, also tracked as ThreeAM, is a Rust-based ransomware run as a ransomware-as-a-service operation in which the core group leases its encryptor to affiliates. First seen in February 2023, it reportedly began as a fallback payload for failed LockBit deployments, and its Russian-speaking operators have been tied to LockBit and BlackSuit. The group steals data, encrypts files with the .threeamtime extension, wipes shadow copies and lists victims on a dark web leak site.

#2

In September 2026, a 3AM affiliate breached a UK school estate of 47 systems spread across several unconnected Active Directory domains via an unknown access vector. Using Impacket-style remote execution, the operator extracted and DPAPI-decrypted credentials from Veeam databases on three backup servers, copied the Chrome saved-password store, exported the SECURITY and SYSTEM hives, ran Mimikatz renamed as sysmon64.exe and created the backdoor domain administrators library and svc_bkpmon.

#3

Reconnaissance was sector-specific, with SYSVOL searched for other schools' names and subnets. The affiliate then built two-way trusts between three domains using dnscmd and netdom and pushed the backdoor account's access to about 40 machines through scheduled tasks. Finance files were staged as Base64 registry values, mirrored with robocopy, zipped and posted to an unmanaged internal host acting as C2. After anti-tampering blocked EDR removal, the host.exe encryptor hit about 37 machines in under half an hour while deleting system-state backups.

#4

The intrusion stands out for its suspected hybrid human and AI operation. Command content strongly indicates language-model authorship, including a Veeam REST call with the wrong port and authentication method, a broken LSASS-dump parser, Unix shell habits in Windows commands, confirmation artifacts such as da_proof.txt and numbered script revisions. Retries arrived every 40 to 101 seconds regardless of change size while new tasks took minutes, consistent with a human setting goals and an agentic loop handling retries, though manual relaying to a chatbot cannot be excluded.

#5

The toolkit is not novel, relying on built-in Windows binaries, public tools and a commodity encryptor. AI involvement changes the traces rather than the playbook, producing repetitive and noisy activity that the controls used against conventional ransomware can still stop.


Recommendations

  1. 01

    Remove Rogue Domain Trusts and Backdoor Admins

    Enumerate every domain trust with nltest /domain_trusts or netdom, remove any trust that lacks an approved change record, and audit Domain Admins and Enterprise Admins in every domain for unexpected members such as library and svc_bkpmon.

  2. 02

    Rotate Credentials Exposed Through Backup Platforms

    Treat every credential stored in Veeam configuration databases, every domain administrator password and saved browser passwords on terminal servers as compromised, rotate them, reset the KRBTGT account twice, and scope backup service accounts to a single domain.

  3. 03

    Alert on Domain Trust Creation

    Raise high-priority alerts for Windows Event ID 4706, netdom trust /add and dnscmd /zoneadd /forwarder executions, since a new inter-domain trust is a rare change in a healthy estate.

  4. 04

    Detect Hive Exports and LSASS Access

    Alert on reg save of the SAM, SECURITY or SYSTEM hives, rundll32 calls to comsvcs.dll MiniDump and remote winreg access from non-administrative hosts, and enable LSA protection and Credential Guard to block credential theft from memory.

  5. 05

    Hunt for Agentic Loop Artifacts

    Search hosts for probe and confirmation artifacts such as da_proof.txt, psx_mark.txt, task_probe.txt and the VEEAMPROBE task, numbered script revisions such as av_remove2.ps1, timestamped output files written to ADMIN$, and bursts of near-identical commands repeated at 40 to 100 second intervals.

  6. 06

    Monitor the Registry as a Data Channel

    Alert on large Base64 string values written under HKLM\SOFTWARE, including VC31, FIN001 to FIN040 and FINCOUNT, and on remote registry writes that originate from unexpected hosts.

  7. 07

    Restrict Remote Execution Paths

    Limit WMI, WinRM, remote registry, remote scheduled-task creation and ADMIN$ or C$ access to designated management hosts, and disable the RemoteRegistry service where it is not required.


Indicators of Compromise (IoCs)

TypeValue
Filenamesav_remove3.ps1grant_lib.ps1hop2.ps1hop3.ps1crackw.pysvcstat.exercv.aspxwmi_test.txt
File PathsC:\Windows\Temp\1\host.exeC:\Windows\Temp\sysmon64.exeC:\Windows\Temp\av_remove.ps1C:\Windows\Temp\av_remove2.ps1C:\Windows\Temp\svc_data.txtC:\Windows\Temp\svc_test.ps1C:\Windows\Temp\prx_go.cmdC:\Windows\Temp\prx.txtC:\Windows\Temp\diag_member.batC:\Windows\Temp\diag_nl.batC:\Windows\Temp\lmhosts_test.batC:\Windows\Temp\wb_probe.batC:\Windows\Temp\nl_restart.batC:\Windows\Temp\da_proof.txtC:\Windows\Temp\task_probe.txtC:\Windows\Temp\psx_mark.txtC:\Windows\Temp\mimi_run.txtC:\Windows\Temp\mimi.logC:\Windows\Temp\ls.dmpC:\Windows\Temp\sec.saveC:\Windows\Temp\sys.saveC:\Windows\Temp\m.zipC:\Windows\Temp\fin.zipC:\Windows\Temp\rst\finance_mirror
URLhxxps[:]//github[.]com/gentilkiwi/mimikatz/releases/download/2[.]2[.]0-20220919/mimikatz_trunk[.]zip
Registry KeysHKLM\SOFTWARE\VC31HKLM\SOFTWARE\FIN001-FIN040HKLM\SOFTWARE\FINCOUNT
Scheduled TaskVEEAMPROBEWindowsTempDiag
Usernamelibrarysvc_bkpmon
File Extension.threeamtime
Recent Breacheshttps://www.fleetworksinc.comhttps://www.stjames.wa.edu.auhttps://www.bhn-expertise.comhttps://www.coosalud.comhttps://www.midwestbit.comhttps://www.apexus.comhttps://www.safescaffolding.nethttps://www.pistonespersan.com.arhttps://www.newmantractor.comhttps://www.wmdn.nethttps://www.mecasem.orghttps://www.clubonecasino.comhttps://www.tws-tac.nethttps://www.guardianbarrierservices.comhttps://www.acemacon.orghttps://www.mgrlaw.comhttps://www.hoplongtech.comhttps://www.agroexportavocados.comhttps://www.amc.org.au

MITRE ATT&CK TTPs

Execution
T1059Command and Scripting Interpreter
T1059.001PowerShell
T1059.003Windows Command Shell
T1059.006Python
T1047Windows Management Instrumentation
T1053Scheduled Task/Job
T1053.005Scheduled Task
T1569System Services
T1569.002Service Execution
Persistence
T1136Create Account
T1136.002Domain Account
T1098Account Manipulation
T1098.007Additional Local or Domain Groups
T1078Valid Accounts
T1078.002Domain Accounts
Privilege Escalation
T1484Domain or Tenant Policy Modification
T1484.002Trust Modification
Defense Evasion
T1562Impair Defenses
T1562.001Disable or Modify Tools
T1562.004Disable or Modify System Firewall
T1036Masquerading
T1036.005Match Legitimate Resource Name or Location
T1036.010Masquerade Account Name
T1112Modify Registry
T1027Obfuscated Files or Information
T1027.010Command Obfuscation
T1140Deobfuscate/Decode Files or Information
T1218System Binary Proxy Execution
T1218.011Rundll32
T1622Debugger Evasion
T1480Execution Guardrails
Credential Access
T1555Credentials from Password Stores
T1555.003Credentials from Web Browsers
T1003OS Credential Dumping
T1003.001LSASS Memory
T1003.002Security Account Manager
T1003.004LSA Secrets
T1003.005Cached Domain Credentials
Discovery
T1482Domain Trust Discovery
T1069Permission Groups
T1069.002Domain Groups
T1033System Owner/User Discovery
T1007System Service Discovery
T1518Software Discovery
T1046Network Service Discovery
T1135Network Share Discovery
T1083File and Directory Discovery
Lateral Movement
T1021Remote Services
T1021.002SMB/Windows Admin Shares
T1021.006Windows Remote Management
T1570Lateral Tool Transfer
Collection
T1039Data from Network Shared Drive
T1074Data Staged
T1074.001Local Data Staging
T1074.002Remote Data Staging
T1560Archive Collected Data
Command and Control
T1105Ingress Tool Transfer
T1071Application Layer Protocol
T1071.001Web Protocols
Exfiltration
T1041Exfiltration Over C2 Channel
Impact
T1486Data Encrypted for Impact
T1490Inhibit System Recovery
T1657Financial Theft

References & Patch Links

References

What Next?

At Hive Pro, it is our mission to detect the most likely threats to your organization and to help you prevent them from happening.

Reduce real exposure. Not just vulnerability volume.