Skip to main content
Mallory
← Back to News
ai-enabled-threat-activitycredential-access-methodlateral-movement-methodconsumer-discretionary-sector-threat

AI Assistance Revives Stalled Intrusion at U.S. Retailer

Emerging Technology Risk · 3 sources · Updated 5d ago · First seen Sep 26, 2026

AI Assistance Revives Stalled Intrusion at U.S. Retailer

Attackers apparently used AI assistance to revive a stalled intrusion at a U.S. retailer, accelerating reconnaissance, credential theft and lateral movement across seven observed computers in September 2026. Symantec assessed AI involvement with high confidence and human oversight with moderate confidence, but could not identify the model or service. The attackers abused SQL Server and compromised web applications, deployed DLL-sideloading loaders and web shells, and harvested credentials and registry hives. Within approximately 10 minutes of examining SolarWinds Orion, they located its RabbitMQ authentication secret and used Erlang remote procedure calls to execute commands on another server. Shared infrastructure suggested a possible connection to another financially motivated retail-targeting operation, but neither that connection nor attribution to a named group was established.

The activity illustrates how suspected AI assistance can accelerate established intrusion techniques rather than introduce novel ones. Separate investigations documented comparable post-exploitation methods: Trend Micro reported Shadowpad activity affecting at least 21 companies across 15 countries, with ransomware deployed in some cases, while Elastic documented REF2924 using DOORME, SIESTAGRAPH and SHADOWPAD for persistence, command-and-control and information collection. Those campaigns included credential dumping, DLL sideloading, registry-stored payloads and abuse of legitimate services; the supplied reporting establishes no connection between them and the retailer intrusion. Defensive priorities include restricting access to application secrets, monitoring unusual cross-server execution and credential dumping, and detecting web shells and DLL sideloading. The retailer case underscores how quickly an exposed application secret can become a lateral-movement path.

New surface, new adversariesTrack how attackers are adapting to this technology.
Track this risk
Event timeline

How this story unfolded

Events from the most recent confirmed update back to the earliest known activity.

27 events · 2022 – Sep 26, 2026
  1. Reported Nov 12, 20265d ago

    Symantec assesses AI assistance in the retailer intrusion

    Symantec assessed AI assistance with high confidence and human-mediated interaction with moderate confidence, citing structured scripts, iterative reasoning and pauses between submissions and strategy changes. It could not identify the model or service, establish attribution to a named threat group, or confirm the suspected link to the separately reported retailer-targeting operation.

    How AI Revived a Stalled Attack on a U.S. Retailer | SECURITY.COM
  2. Reported Nov 12, 20265d ago

    Orion RabbitMQ secret enables command execution on another retailer server

    Within approximately 10 minutes of first querying Orion services, attackers located RabbitMQ's .erlang.cookie secret and used Orion's erl.exe to execute commands on a second web server through Erlang remote procedure calls. Commands used decimal character-code encoding and queried account information, cached logons and LSA secrets.

    How AI Revived a Stalled Attack on a U.S. Retailer | SECURITY.COM
  3. Reported Nov 12, 20265d ago

    Retailer attackers harvest application credentials and inspect monitoring systems

    Credential searches targeted connection strings, FileZilla Server and client settings, WinSCP, PuTTY, saved RDP files, IIS configuration and Windows logon registry values. Attackers also queried SolarWinds Orion's database, read a Zabbix agent pre-shared key and tested credentials for sysadmin privileges on two remote SQL Servers.

    How AI Revived a Stalled Attack on a U.S. Retailer | SECURITY.COM
  4. Reported Nov 12, 20265d ago

    Retailer attackers adapt launch methods after security-product blocks

    Following repeated security-product terminations, attackers used misleading executable names and BreakChain's WMI and shell modes to alter process ancestry. A second BreakChain build appeared approximately 75 minutes after the first was blocked, and ask.exe was also renamed w3wp.exe to resemble an IIS worker process.

    How AI Revived a Stalled Attack on a U.S. Retailer | SECURITY.COM
  5. Reported Nov 12, 20265d ago

    Reverse shell and web shells provide additional retailer access

    Attackers retrieved server.ps1, an AES-encrypted PowerShell reverse shell connecting to 166.0.25[.]106:443, using a fallback download host when needed. They also placed filemanager.asp and qwdajhsd.asp in the web application's images directory to retain access independently of the original entry point.

    How AI Revived a Stalled Attack on a U.S. Retailer | SECURITY.COM
  6. Reported Nov 12, 20265d ago

    Retailer attackers deploy DLL-side-loading payloads

    Attackers staged Bitdefender's bdreinit.exe with malicious log.dll and encrypted log.txt, later using the G DATA-signed ask.exe with ask.dll and ask.txt. Symantec cautioned that the Bitdefender pairing alone did not establish use of ShadowPad or identify a threat actor.

    How AI Revived a Stalled Attack on a U.S. Retailer | SECURITY.COM
  7. Reported Nov 12, 20265d ago

    Retailer attackers reconstruct EfsPotato and collect registry secrets

    On Computer 2, attackers reconstructed the publicly available EfsPotato exploit from base64-encoded material, decoded it with Certutil and compiled it with csc.exe to obtain SYSTEM-level execution. They also saved SAM, SYSTEM and SECURITY registry hives containing potentially valuable credential material.

    How AI Revived a Stalled Attack on a U.S. Retailer | SECURITY.COM
  8. Reported Nov 12, 20265d ago

    Three additional South Asian intrusions linked through infrastructure or tools

    Symantec linked three additional South Asian cases through shared infrastructure or tooling. The activity involved persistent backdoors, privilege escalation, password theft, tunnelling and spreadsheet exfiltration.

    How AI Revived a Stalled Attack on a U.S. Retailer | SECURITY.COM
  9. Sep 26, 20262mo ago

    Password spraying begins from a seventh retailer computer

    Computer 7 began testing downloaded credential pairs against a web service. The password-spraying activity continued until at least September 29.

    How AI Revived a Stalled Attack on a U.S. Retailer | SECURITY.COM
  10. Sep 24, 20262mo ago

    Retailer attackers expand credential collection and internal reconnaissance

    From September 24 onward, attackers dumped additional registry hives, used Chisel, downloaded sqlrun2.ps1 and searched for NTDS.dit, Group Policy Preferences cpassword values and LAPS passwords. They also tested SQL sa passwords and scanned more than a dozen internal hosts for SQL Server, SMB, RDP and WinRM services.

    How AI Revived a Stalled Attack on a U.S. Retailer | SECURITY.COM
  11. Sep 23, 20262mo ago

    Retailer attackers switch to longer, structured scripts

    Attackers submitted approximately 50 scripts over roughly two hours, with a median length exceeding 500 characters, no typing errors, labelled output, error handling and output limits. Comments referenced previous results and revised hypotheses, providing strong evidence of AI assistance in Symantec's assessment.

    How AI Revived a Stalled Attack on a U.S. Retailer | SECURITY.COM
  12. Sep 22, 20262mo ago

    Gambit Security identifies overlapping retailer-targeting infrastructure

    Gambit Security identified 155.254.22[.]215 as infrastructure and an AI console for a financially motivated operation targeting retailers. Symantec later noted a possible connection through this shared address and point-of-sale credential searches, without establishing that the operations were linked.

    How AI Revived a Stalled Attack on a U.S. Retailer | SECURITY.COM
  13. Sep 22, 20262mo ago

    Retailer web-server activity shows repeated loader failures

    Attackers performed 110 distinct actions averaging approximately 45 characters on the compromised web server. The activity included three quickly corrected typing errors and five unsuccessful attempts to execute the same blocked loader.

    How AI Revived a Stalled Attack on a U.S. Retailer | SECURITY.COM
  14. Sep 21, 20262mo ago

    Retailer intrusion reaches an IIS web server

    The retailer's Computer 4 first showed malicious activity, with commands executing beneath the IIS worker process w3wp.exe.

    How AI Revived a Stalled Attack on a U.S. Retailer | SECURITY.COM
  15. Sep 18, 20262mo ago

    Attackers execute commands on three SQL Servers at a U.S. retailer

    Attackers executed commands through cmd.exe beneath sqlservr.exe, most likely using xp_cmdshell. Early scripts searched point-of-sale synchronization directories for credentials and connection strings and tested connectivity to 155.254.22[.]215:9458.

    How AI Revived a Stalled Attack on a U.S. Retailer | SECURITY.COM
  16. Aug 2026

    East African ICT provider suffers suspected AI-assisted intrusion

    Attackers spent approximately 61 hours inside an ICT provider serving banks and telecommunications companies, targeting Active Directory credentials and mobile-payment integration secrets. During a suspected AI-assisted phase, they submitted 212 structured scripts averaging approximately 1,000 characters.

    How AI Revived a Stalled Attack on a U.S. Retailer | SECURITY.COM
  17. Aug 2026

    Separate intrusion uses infrastructure later seen in the retailer attack

    An intrusion at another organization used the Yuze tunnelling tool to contact 204.194.49[.]129. The same address later served payloads during the U.S. retailer intrusion, although the infrastructure overlap did not establish attribution.

    How AI Revived a Stalled Attack on a U.S. Retailer | SECURITY.COM
  18. Aug 2026

    South Asian government hydroelectric company targeted

    Attackers targeted database credentials and lateral movement at a South Asian government hydroelectric company in an intrusion involving suspected AI assistance.

    How AI Revived a Stalled Attack on a U.S. Retailer | SECURITY.COM
  19. Jul 2026

    Attackers compromise a South Asian shared-hosting server

    An intrusion involved roughly 970 commands, database credential theft, scheduled-task persistence and attempted exploitation of CVE-2022-21882. Symantec identified it as another suspected AI-assisted incident.

    How AI Revived a Stalled Attack on a U.S. Retailer | SECURITY.COM
  20. Jun 2026

    Vietnamese university web server compromised in suspected AI-assisted attack

    Attackers altered application files and timestamps on a Vietnamese university web server and reconstructed PrintSpoofer after six failed download attempts. Chinese-language operator notes embedded in a script supported suspected AI assistance.

    How AI Revived a Stalled Attack on a U.S. Retailer | SECURITY.COM
  21. Reported Feb 20, 20252y ago

    Shadowpad investigation uncovers ransomware and a multinational campaign

    Trend Micro and Orange Cyberdefense CERT linked the investigated activity to at least 21 companies across 15 countries and found a previously unreported ransomware family deployed alongside Shadowpad in some incidents. Analysis revealed updated Shadowpad anti-debugging, registry-based payload encryption and DNS-over-HTTPS capabilities; attribution remained uncertain, with only weak links to Teleboyi.

    Updated Shadowpad Malware Leads to Ransomware Deployment
  22. Nov 2024

    Shadowpad found in two European incident-response cases

    Two European incident-response cases revealed Shadowpad infections. The investigated activity involved remote network access, weak passwords and bypasses of multifactor authentication.

    Updated Shadowpad Malware Leads to Ransomware Deployment
  23. Reported Feb 7, 20234y ago

    Elastic documents REF2924 backdoors and assesses campaign relationships

    Elastic documented DOORME, SIESTAGRAPH and a SHADOWPAD loader found on two domain controllers that used an outdated Bitdefender executable for DLL side-loading and stored its payload in the registry. It assessed with moderate confidence that the related campaigns involved a nationally aligned, non-monetary threat group with overlaps to Winnti and ChamelGang activity.

    Update to the REF2924 intrusion set and related campaigns
  24. Reported Feb 7, 20234y ago

    Matching DOORME implant links a second Afghan telecom to REF2924

    Elastic identified an identically configured DOORME backdoor on an internet-connected Exchange server at a second telecommunications provider in Afghanistan. It associated that activity with REF2924 with high confidence based on the configuration, public Exchange-server scanning and shared strategic targeting.

    Update to the REF2924 intrusion set and related campaigns
  25. Dec 2022

    REF2924 targets an ASEAN foreign ministry and collects Exchange mailboxes

    Elastic observed DOORME targeting the Foreign Affairs Office of an ASEAN member state. In early December, attackers used PowerShell to collect and export mailboxes from the organization's internet-connected Exchange server.

    Update to the REF2924 intrusion set and related campaigns
  26. Nov 1, 20224y ago

    Microsoft Graph refresh token generated for SIESTAGRAPH

    The refresh token embedded in the analyzed SIESTAGRAPH backdoor was generated on November 1. The implant used Microsoft Graph access to Outlook drafts and OneDrive for command-and-control and file transfers.

    Update to the REF2924 intrusion set and related campaigns
  27. 2022

    Winnti-linked intrusion observed at an Afghan telecommunications provider

    Elastic observed a short-lived intrusion at a telecommunications provider in Afghanistan that it separately tracked as Winnti-linked activity.

    Update to the REF2924 intrusion set and related campaigns
Linked entities

Related entities

Vulnerabilities, threat actors, malware, products, organizations, breaches, and observables Mallory has linked to this story. Indicator values are masked here and available in full in the app.

Open in app →

Affected products

2 linked
Internet Information ServicesSql Server

Observables

93 linked

Indicator values are masked on this page. View all 93 in Mallory Domains, IPs, hashes, and URLs are exportable to your SIEM.

Do you run this?Follow how adversaries are adapting to this technology, and where it touches your stack today.
Track This Risk
Mallory platform

See the full picture, correlated to your attack surface.

This page covers what’s public. Mallory adds the parts that aren’t: which of your assets are affected, which threat actors are using it right now, and what to deploy next.

Exposure mappingMap indicators from this story to your assets and identify affected systems in minutes.
Threat actor evidenceEvery observed campaign, victim, and pivot linked to actors named in this story.
Associated malwareMalware, exploits, and IOCs connected to the activity described here.
Detection signaturesYARA, Sigma, and Snort rules deployed to your SIEM as soon as they’re published.
Scheduled alertsGet matching new stories delivered to your team as they break, not the next morning.
AI threadsAsk questions about this story and take action on the answers.