Threat reportThreat IntelligenceTL-2026-6454
Attackers Use AI-Powered ARTEX Vulnerability-Discovery Framework in Cyberattack on European Commission IT Systems
CERT-EU detected a cyberattack on European Commission IT systems in early November 2026, reported to involve ARTEX, a publicly available LLM-driven penetration-testing framework that the FBI warned on 2026-11-13 is being weaponized by multiple actors. The Commission took applications such as Sysper, MyPMO and translation services offline for more than 10 days; whether staff data was accessed is undetermined and no attacker has been named.
Key facts
| Identifiers | TL-2026-6454 |
|---|---|
| Severity | HIGH |
| First disclosed | |
| Affected products | European Commission Sysper (HR platform), MyPMO (medical reimbursement), translation services and other internal IT applications |
| Tools | ARTEX, ARTEX PRO |
| MITRE ATT&CK techniques | T1078.004 Cloud Accounts, T1087.002 Domain Account, T1110.003 Password Spraying, T1190 Exploit Public-Facing Application, T1199 Trusted Relationship, T1505.003 Web Shell, T1552.001 Credentials In Files, T1588.007 Artificial Intelligence, T1595.002 Vulnerability Scanning |
| Motivation | UNKNOWN |
| Target sectors | government administration, european-institutions, finance, technology, defense, health, telecoms, legal |
| Target regions | Europe |
| Detection rules | 9 (1 shown free) |
| Sources | 9 cited sources |
Detection coverage
As of 2026-11-19, Threadlinqs Intelligence publishes 9 detection rule(s) for TL-2026-6454 across Splunk SPL, Microsoft KQL and Sigma, covering 8 indicator(s) of compromise. The whole corpus is readable without an account. Every threat comes with one complete detection in Splunk SPL, Microsoft KQL and Sigma, free for everyone; a free account adds daily TL Creds to unlock the rest. Paid plans add a larger daily allowance, correlation and the MCP server. Threadlinqs MCP server · View plans.
Free detection: ARTEX-Style Automated Vulnerability Scanning of Public Web Portals (TL-2026-6454)
Detects a single client probing a large number of distinct URIs on an IIS-hosted public application within 10 minutes while producing a high share of 4xx/5xx responses. This is the footprint of automated AI-agent reconnaissance and vulnerability discovery such as the ARTEX framework (FBI FLASH 2026-11-13) used against overlooked employee and partner portals like Sysper and MyPMO. Use of ARTEX in the European Commission incident itself is not confirmed; the rule is behaviour-first and does not depend on any published IOC.
- Languages: Splunk SPL, Microsoft KQL, Sigma
- ATT&CK: T1595.002, T1190
- Data sources: iis_access_logs (index=web, sourcetype=ms:iis:auto), W3CIISLog, webserver logsource
- Severity: high
- Confidence: high
Splunk SPL
```Flags one client touching many distinct URIs with a high error ratio which is typical of automated agent scanning```
index=web sourcetype=ms:iis:auto earliest=-1h
| bin _time span=10m
| stats count AS requests dc(cs_uri_stem) AS distinct_uris dc(cs_uri_query) AS distinct_queries sum(eval(if(sc_status>=400,1,0))) AS error_responses values(cs_User_Agent) AS user_agents by _time c_ip s_sitename
| eval error_ratio=round(error_responses/requests,2)
| where requests>=300 AND distinct_uris>=100 AND error_ratio>=0.3
| eval threat="TL-2026-6454 ARTEX automated vulnerability scanning", mitre="T1595.002,T1190"
| table _time c_ip s_sitename requests distinct_uris distinct_queries error_ratio user_agents threat mitreMicrosoft KQL
// ARTEX-style automated scanning of public IIS applications (TL-2026-6454)
W3CIISLog
| where TimeGenerated > ago(24h)
| summarize Requests = count(), DistinctUris = dcount(csUriStem), DistinctQueries = dcount(csUriQuery), ErrorResponses = countif(toint(scStatus) >= 400), UserAgents = make_set(csUserAgent, 5) by bin(TimeGenerated, 10m), cIP, sSiteName
| extend ErrorRatio = round(todouble(ErrorResponses) / Requests, 2)
| where Requests >= 300 and DistinctUris >= 100 and ErrorRatio >= 0.3
| project TimeGenerated, cIP, sSiteName, Requests, DistinctUris, DistinctQueries, ErrorRatio, UserAgentsSigma
title: Web Server Error Response Used By Automated Scanning (ARTEX Correlation Base)
id: 6f1c3a52-8b1e-4d27-9a43-1c5e7d2b9a01
name: tl6454_web_scan_errors
status: experimental
description: Base rule matching client error responses from a web server. Used by the correlation rule to detect automated vulnerability scanning associated with ARTEX (TL-2026-6454).
references:
- https://www.ic3.gov/CSA/2026/261113.pdf
- https://gbhackers.com/hackers-exploit-ai-powered-artex-tool/
author: Threadlinqs Team
date: 2026-11-19
tags:
- attack.reconnaissance
- attack.t1595.002
- attack.initial-access
- attack.t1190
logsource:
category: webserver
detection:
selection:
sc-status:
- 400
- 401
- 403
- 404
- 405
condition: selection
falsepositives:
- Legitimate vulnerability scanners
level: informational
---
title: ARTEX-Style Automated Web Vulnerability Scanning (TL-2026-6454)
id: 2b8d4e61-3c7a-4f95-8d10-5a9e6c3f7b02
status: experimental
description: Detects one client generating error responses across 100 or more distinct URIs within 10 minutes, consistent with AI-agent driven vulnerability scanning of public portals (ARTEX, TL-2026-6454).
references:
- https://www.ic3.gov/CSA/2026/261113.pdf
- https://gbhackers.com/hackers-exploit-ai-powered-artex-tool/
author: Threadlinqs Team
date: 2026-11-19
tags:
- attack.reconnaissance
- attack.t1595.002
- attack.initial-access
- attack.t1190
correlation:
type: value_count
rules:
- tl6454_web_scan_errors
group-by:
- c-ip
timespan: 10m
condition:
gte: 100
field: cs-uri-stem
falsepositives:
- Authorized vulnerability scanners and external attack-surface monitoring
- Broken crawlers or misconfigured health checks
level: highKnown false positives:
- Authorized vulnerability scanners and external attack-surface management services
- Search-engine or monitoring crawlers that request many missing URLs
- Load tests against a staging portal exposed through the same site
Tuning notes: Allow-list internal scanner and ASM source IPs. Raise or lower the 300 request / 100 distinct URI / 0.3 error-ratio thresholds per site baseline. Point the search at the index and sourcetype that hold your reverse-proxy or WAF logs if IIS is not the front end. Prioritise HR, reimbursement and partner portals.
- Password Spraying Against Entra ID Accounts From a Single Source (ARTEX-Linked) · unlock with TL Creds
- Service Principal Sign-In From New Source Infrastructure (Vendor App Abuse) · unlock with TL Creds
- Rapid Active Directory Account and Group Enumeration From a Single Host · unlock with TL Creds
- ARTEX or ARTEX PRO Framework Binary Execution · unlock with TL Creds
- Credential Hunting in Files via Command-Line Search Utilities · unlock with TL Creds
- Sequential Record Lookups by a Single Session on Employee or Partner Portals · unlock with TL Creds
- Web Shell Written to Web Root by Web Server Process (ARTEX-Linked Persistence) · unlock with TL Creds
- Non-Browser Process Resolving LLM API Endpoints (ARTEX Agent Egress) · unlock with TL Creds
Indicators of compromise
As of 2026-11-19, Threadlinqs Intelligence tracks 8 indicators of compromise for TL-2026-6454.
Technical analysis
In early November 2026, CERT-EU, the cybersecurity service for EU institutions, detected malicious activity against European Commission IT systems (first reported via POLITICO's Brussels Playbook and relayed by GBHackers on 2026-11-19). As a precaution the Commission took several applications offline, including Sysper (the HR platform used for payslips, leave requests and timesheets), MyPMO (medical reimbursements) and translation services. The disruption lasted more than 10 days; Sysper and MyPMO were restored the week of 16 November and Sysper briefly went down again on 18 November. Salaries were processed normally but some staff could not submit timesheets or request leave. Commission spokesperson Thomas Regnier confirmed that individual systems were targeted and that those concerned were informed. An urgent Corporate Management Board meeting was held, DG DIGIT added security checks, and Commissioner Henna Virkkunen briefed the College on 18 November. An EU official told POLITICO the attack was believed to be largely contained but its full extent was unclear. Precautionary shutdowns do not imply that every affected application was compromised.
The reporting links the activity to ARTEX, described as an open-source, Go-based autonomous penetration-testing framework that uses AI agents to coordinate reconnaissance, vulnerability discovery, exploitation, privilege escalation and data theft with little human input. ARTEX has no model of its own; it integrates external LLMs (Anthropic Claude, OpenAI GPT, DeepSeek, GLM, Grok, Kimi). It was published on GitHub in July 2026 by a developer using the handle Autumn-27, the repository was removed on 2026-10-08 after abuse reports, and the framework persists through forks, mirrors, archived copies, localized variants and a commercial edition, ARTEX PRO. Importantly, the source material does not establish that ARTEX was the means of access in the Commission incident, does not identify the vulnerability or initial access vector, and states that identifying the tool does not identify the operator because the same framework appears in unrelated campaigns.
Context from the FBI FLASH alert (published 2026-11-13, linked as ic3.gov/CSA/2026/261113.pdf; the identifier is inferred from the URL) and follow-on reporting: multiple actors have weaponized ARTEX since at least August 2026 against financial, government, technology, defense, healthcare, telecom, legal and other sectors, with some intrusions reaching privileged access in under three hours. Reported ARTEX-linked campaigns include the October 2026 ConnectWise Cloud Backup (formerly SkyKick) supply-chain compromise (attributed by Microsoft to China-aligned Storm-0940; password spraying against 327 accounts at eight organizations produced over 151,000 failed sign-ins and later app-only authentication using compromised ConnectWise application identities), intrusions at South Korean banks (about 67,000 people's data), the Barracuda breach (an unrevoked third-party credential), and the CrowdStrike-reported Vapor Panda cyberespionage campaign (web shells, persistence). Researchers counted more than 1,600 publicly exposed ARTEX consoles in early November. None of these campaigns has been linked to the Commission incident. Earlier 2026 Commission incidents (January mobile-device-management intrusion; March AWS account breach tied by CERT-EU to TeamPCP via the Trivy supply-chain compromise) are separate and not reported as connected.
Technique mapping below reflects ARTEX-associated TTPs documented by the FBI and vendors and is not confirmed Commission-specific tradecraft. No IOCs (IPs, domains, hashes) were published for the Commission incident; the IOC list therefore contains tool, actor-context and behavioral indicators. No network IOCs were available to check against BeaconBeagle.
MITRE ATT&CK techniques by tactic:
Initial Access
T1078.004 Cloud Accounts; T1190 Exploit Public-Facing Application; T1199 Trusted Relationship
Discovery
Credential Access
T1110.003 Password Spraying; T1552.001 Credentials In Files
Persistence
Resource Development
T1588.007 Artificial Intelligence
Reconnaissance
Remediation
Patches
- No CVE or vendor patch has been identified for this incident; patch internet-facing software promptly as general hygiene
Immediate actions
- Inventory all internet-facing applications (partner portals, employee apps, mobile backends, legacy services) and assign named owners
- Review service-principal and vendor-app sign-ins separately from user sign-ins; flag sign-ins from outside expected vendor infrastructure
- Rotate credentials where access cannot be ruled out and revoke third-party credentials when work ends
- Keep administrative interfaces and AI tooling off the public internet and require authentication
Workarounds
- Monitor outbound connections to LLM API endpoints or self-hosted LLM gateways and unexplained AI API billing spikes
- Apply the MalwareMustDie open-source ARTEX detection pack, treating infrastructure indicators as potentially stale
- Report suspicious activity to the FBI (IC3) or national CERT
Longer-term hardening
- Enforce MFA and least privilege on multi-tenant application permissions and consent grants; use just-in-time elevation
- Scan repositories, CI/CD, IaC and container images for embedded credentials and remove hardcoded secrets
- Rate-limit partner and employee portals; alert on sequential lookups and abnormal record counts per session
- Run recurring authentication and authorization tests on internet-facing applications
- Retain detailed logs sufficient to rule access in or out
Affected and fixed versions
- European Commission — Sysper (HR platform), MyPMO (medical reimbursement), translation services and other internal IT applications
Vulnerable versions: Not disclosed
Timeline
- ARTEX, an open-source Go-based AI penetration-testing framework, is first published on GitHub by developer 'Autumn-27' (July 2026; exact day not reported).
- Earliest documented weaponization of ARTEX by multiple threat actors, per the FBI (August 2026; exact day not reported).
- Developer removes the main ARTEX GitHub repository after reports of abuse; forks, archived copies and ARTEX PRO remain in circulation.
- Microsoft publishes its investigation of Storm-0940 and the ARTEX-linked ConnectWise Cloud Backup supply-chain compromise.
- Researchers report hundreds of exposed ARTEX systems; over 1,600 public consoles counted by early November.
- CERT-EU detects the attack on European Commission IT systems in early November 2026 (exact day not published; date approximate); the Commission takes applications offline as a precaution.
- Barracuda publishes its incident analysis of an ARTEX-linked breach via an unrevoked third-party credential.
- FBI publishes a FLASH alert on weaponized ARTEX; MalwareMustDie releases an open-source ARTEX detection pack.
- Sysper and MyPMO resume earlier in the week of 16 November after more than 10 days of disruption (approximate).
- Commissioner Henna Virkkunen briefs the College of Commissioners; Sysper briefly goes offline again.
- GBHackers, citing POLITICO, reports the ARTEX-linked Commission attack; most applications are back online, data-theft status undetermined, no attacker named.
Sources cited
- FBI FLASH on ARTEX (IC3 PDF 261113) ic3.gov · Advisory · 2026-11-13
- FBI ARTEX Alert: AI Attacks Are Logging in, Not Breaking In (SecureWorld) secureworld.io · Research · 2026-11-16
- Hackers Exploit AI-Powered ARTEX Tool to Identify Vulnerabilities in European Commission IT Systems gbhackers.com · News · 2026-11-19
- POLITICO Brussels Playbook: Cyberattack hits the Commission politico.eu · News · 2026-11-19
- FBI warns ARTEX AI hacking tool is powering global cyberattacks (BleepingComputer) BleepingComputer · News · 2026-11-15
- FBI says ARTEX AI open source hacking tool used to breach networks in under three hours (TechCrunch) techcrunch.com · News · 2026-11-16
- FBI says AI hacking tool ARTEX can break into networks in under three hours (The Next Web) thenextweb.com · News · 2026-11-16
- ARTEX AI tool FBI flash alert (The Record) The Record · News · 2026-11-16
- FBI warns ARTEX AI framework (GBHackers) gbhackers.com · News · 2026-11-16
Drafted by the Threadlinqs research pipeline from 9 sources; published ; last updated —.