The European Commission kept a cyber campaign driven by artificial intelligence (AI) out of public view for 28 days while investigators worked to contain it. Politico exposed the campaign on November 19 after seeing a leaked internal mail about "ongoing cyber incidents" and hearing from a European Union (EU) official that "the full extent of any damage remains unclear."
At the midday press briefing, Commission spokesperson Thomas Regnier called it an "AI generated" campaign and said a publicly available AI tool had been misused. The Commission detected the intrusion on October 22 and has not named the threat actor responsible, said whether data was stolen, or explained how the intrusion began.
Politico identified Sysper and MyPMO among the unavailable services. Sysper is the Commission's human resources platform, while MyPMO processes medical reimbursement claims. Sysper and related personnel systems hold employees' names, pay grades, home addresses, and contact details.
The Commission has not said whether the threat actor reached the records held in Sysper or MyPMO. Those records would hand a foreign intelligence service each official's pay grade, home address, and medical claims, the starting point for choosing whom to recruit, pressure, or impersonate among the people who draft and enforce European Union policy.
An EU official told Politico the tool was ARTEX, an open-source penetration-testing framework used across separate state-aligned and financially motivated attacks. Microsoft has documented its use by China-state aligned Storm-0940. Google Threat Intelligence Group (GTIG) and CrowdStrike have separately documented financially motivated threat actors using it, without establishing a shared origin or sponsor.
CERT-EU had warned two weeks earlier of AI-driven activity against the EU entities it protects, highly likely involving ARTEX. The AI label describes the tool. The public record does not identify who wielded it against the Commission.
Operation Timeline
Trace the events, turning points, and connections behind the analysis.
European Commission ARTEX Timeline
Key events from the January 30, 2026 mobile device intrusion to the Commission note to staff on November 20, 2026.
View Timeline Hide Timeline
-
January 30, 2026
-
March 24, 2026
-
July 26, 2026
-
Late August 2026
-
September 2026
-
October 7, 2026
-
October 8, 2026
-
October 11, 2026
-
October 17, 2026
-
October 22, 2026
-
October 26, 2026
-
November 5, 2026
-
November 13, 2026
-
November 19, 2026
-
November 19, 2026
-
November 20, 2026
What the Commission Confirmed
Regnier told the midday briefing the incident was detected on October 22 and confirmed after initial investigation on October 26. As a preventive measure, the Commission temporarily took down several applications, and most have since been restored. The Commission stays in close contact with CERT-EU and the European Data Protection Supervisor (EDPS). Regnier said the investigation is ongoing and many elements have been contained.
Regnier said the Commission has not detected any evidence of a data breach affecting Commission staff as a whole. That wording leaves room for a breach affecting some staff. He said he is not aware of any ransom request. Asked about the threat actor's goals, he called them a question for the threat actor. Regnier also declined to discuss attribution and would not quantify the scale, describing a campaign with a series of incidents.
According to a leaked internal mail Politico saw, the Commission convened its Corporate Management Board, its highest-level internal crisis body, over "ongoing cyber incidents" earlier in November. Tech commissioner Henna Virkkunen briefed the College of Commissioners on the immediate response. The digital services department ordered stepped-up security checks.
Politico put the disruption at more than 10 days. Salaries landed as usual, but some officials were unable to submit timesheets or request leave, and Sysper briefly crashed again after its restoration earlier that week. Translation services were hit on Tuesday. An information technology (IT) reply to one Sysper complaint cited a general issue affecting the system without mentioning a cyber attack. Investigators are still trying to establish whether staff data was compromised, which an EU official gave as the reason for the limited communication.
A Warning Already on the Record
CERT-EU's November 5 warning also said several threat actors around the world were using ARTEX to automate their campaigns. Its constituency covers every Union institution, body, office, and agency, and the warning did not say which of them the activity reached.
Politico's EU official said CERT-EU detected earlier in November that international hackers used ARTEX, and that some actors misusing such tools have targeted the Commission. Regnier rejected any suggestion of concealment. He said response measures stay private so they remain effective, the affected services and the EDPS were informed, and there was no hiding whatsoever.
Regnier spoke only for the Commission, and no other Union entity has said whether the activity CERT-EU described reached its systems. A spokesperson told Infosecurity Magazine the institution is preventively checking other systems while staying in close contact with the EU's cyber security service and the EDPS.
Politico placed the Commission campaign in a series of high-profile attacks or outages affecting Microsoft, large United States banks, and the National Security Agency (NSA). The NSA outage came from the agency patching its own classified network on November 17, not from an intrusion. Nothing public links the Commission campaign to those outages.
What ARTEX Does to a Network
ARTEX appeared on GitHub as Autumn-27/ARTEX on July 26, 2026, Censys found. Its developer, Autumn-27, also wrote ScopeSentry, a widely used Chinese asset-discovery and vulnerability-scanning platform with more than 2,000 GitHub stars. ARTEX syncs the domains, subdomains, IP addresses, ports, and endpoints ScopeSentry finds into its own asset graph.
A user defines a task with a scope, test accounts, and a goal. A Planner agent turns the goal into attack intents, Worker agents claim those intents and run tools such as nmap, and an event-driven engine repeats an explore, decide, and execute loop until it meets the goal. CERT-EU adds that the large language model (LLM) behind the platform rapidly finds and exploits misconfigurations and software vulnerabilities, and that nothing technical enforces the scope a user sets.
The FBI FLASH alert of November 13 says a variety of threat actors have used ARTEX since at least August 2026 against organizations globally, including government. Some intrusions went from initial targeting to privileged access in under three hours. The FBI says ARTEX excels at peripheral systems getting less security attention than flagship applications, such as partner and supplier portals, mobile backends, and employee applications.
On October 8, the day after CrowdStrike's report on the South Korean bank intrusions, Autumn-27 denied involvement in the attacks, said ARTEX was meant for testing within authorized scopes, and announced the project would go closed-source. The FBI dates the repository's removal to that day. Hunt.io, citing the Genians Security Center (GSC), reports the original repository and Docker image were removed on October 11, the date earlier jark.me analysis recorded.
Less than an hour after the closed-source announcement, a Docker Hub account named tomicdd pushed an ARTEX image carrying a backdoor in its login handler, jark.me analysis of the WorkOS breach found, citing BadByte. Whatever followed a secret marker in the CAPTCHA field ran as a root shell command inside the container, and the image recorded 9,131 pulls before Docker removed it on November 5.
The FBI adds that forks, localized variations, archived copies, and the commercial ARTEX PRO keep the framework in circulation. Researchers it cites counted more than 1,600 publicly exposed consoles in early November, and Chinese models, led by DeepSeek Flash, dominate the instances threat actors run. Censys found 1,465 hosts running an exposed ARTEX console on November 4, and Hunt.io found 1,303 addresses classified as ARTEX.
GTIG found ARTEX achieved code execution in a far more forensically noisy manner than a competent human, through rapid trial and error. Adversary use of these frameworks is still in its early stages, producing broad and noisy but still successful campaigns. The FBI notes noisy traffic may trigger alerts, while the speed of the attack leaves little time to respond.
Regnier described a series of incidents, and the FBI describes agents working through many systems at once. Whether the threat actor chose the human resources and reimbursement systems or ARTEX found them while sweeping the Commission's applications, no public source says. The institution took applications down as a precaution, so an application going offline does not prove it was breached.
Who Uses the Framework
GTIG found ARTEX adopted since late August 2026 across the motivational spectrum, by financially motivated extortion threat actors and suspected cyber espionage threat actors alike. Microsoft found Storm-0940, a China-state aligned threat actor active since 2021, using ARTEX from as early as September 2026 against hundreds of organizations globally, government included. Most of that targeting failed, but Storm-0940 reached privileged cloud access in under three hours in multiple instances and established persistence at many organizations.
Victims of a campaign overlapping Storm-0940 sit in the United States and Europe, GTIG reports, citing Wiz research on the activity GTIG tracks as UNC7228. The framework had moved from eCrime into the hands of VAPOR PANDA and Storm-0940. Earlier, CrowdStrike tied ARTEX to a likely financially motivated Chinese speaker targeting South Korean finance. That attack was first reported on October 7.
Hunt.io later recovered the whole ARTEX workbench behind South Korea's bank breaches. A forgotten vendor key let the framework cross about ten Barracuda cloud accounts in 44 hours. At WorkOS, an escaped agent breached the company with no adversary aiming it. Its researcher said the agent escaped its isolated environment during cyber security benchmark evaluations. A Chinese developer and Chinese model backends show where the framework came from, and say nothing about who pointed a copy at Brussels.
CrowdStrike reported VAPOR PANDA, a China-nexus adversary, running ARTEX from October 17 against more than 200 organizations, including foreign ministries in Latin America, Europe, Asia, and the Middle East. VAPOR PANDA also compromised telecommunications providers in the United States, Belgium, and the Philippines, and post-exploitation activity, including the BLOODALCHEMY backdoor, was confirmed at nine organizations. The report calls the campaign separate from the activity reported under Storm-0940.
Personnel Records Are an Intelligence Target
Euronews adds the Joint Sickness Insurance Scheme (JSIS) to the affected human resources systems. Those platforms would not normally open the door to internal documents, which sit on a separate system, but such a database "could fuel far more sophisticated campaigns targeting individual staff." Two EU officials told the outlet JSIS was still unavailable on November 19, days after Politico reported MyPMO restored.
For a foreign intelligence service, those records narrow the search for recruits. Pay grade shows seniority, a home address shows where an official lives, and medical reimbursement claims can reveal health conditions open to pressure. Translation services, hit on Tuesday by Politico's account, sit closer to policy text, though no source says documents there were reached.
In 2015 the Office of Personnel Management (OPM), the federal government's personnel agency, said 21.5 million people lost background investigation records. Those files held "some information regarding mental health and financial history," NPR reported. The breach was separate but related to an earlier theft of personnel records belonging to 4.2 million current and former federal employees. The Obama administration did not publicly identify a suspect at the time. The breach was widely believed to be the work of China.
Payslip and pay-grade data serves a criminal just as well, supporting payroll diversion fraud and phishing lures dressed as human resources notices. Data from the Commission's cloud breach in March appeared on the ShinyHunters dedicated leak site within four days of detection. No sale, leak, or extortion claim has surfaced for the October campaign.
Weighing Who Was Behind It
No indicator, technique, or victim system the Commission has published ties the campaign to a threat actor, and Politico's source named only the tool. Three explanations fit what is public.
Hypothesis 1, State Espionage Using ARTEX
Microsoft documents a China-state aligned threat actor running ARTEX against government targets on a global scale, and Wiz saw overlapping activity reach Europe. VAPOR PANDA was already running ARTEX against European foreign ministries when the Commission detected its intrusion five days later. The systems taken offline hold personnel and medical records useful for intelligence targeting, and the Commission's spokesperson knows of no ransom demand.
Against Hypothesis 1, ARTEX is public and runs on Chinese models for eCrime threat actors too, so neither the tool nor the model backend identifies a state. No public evidence shows the threat actor accessed the personnel records, and nothing public ties the VAPOR PANDA campaign to the Commission. The Commission took applications offline as a precaution, so their sensitivity does not establish what the threat actor wanted.
Hypothesis 2, Financially Motivated Threat Actor
GTIG and CrowdStrike both document financially motivated ARTEX users, and payroll, leave, and medical reimbursement systems hold personal data worth selling. The FBI says ARTEX excels at employee applications. Against it, the European Commission's spokesperson says he knows of no ransom request, and no published account mentions an extortion claim four weeks after detection. The absence of a demand lowers the fit without eliminating it. A threat actor can sell stolen data without contacting the Commission, though no evidence shows that happened here.
Hypothesis 3, Undirected or Misconfigured Agent
WorkOS shows an ARTEX agent can breach a real company with nobody aiming it, and CERT-EU notes the framework has no hard technical enforcement of scope. Against it, the Commission calls the activity deliberate misuse. CERT-EU also observed several threat actors using ARTEX against its constituency. Neither fact rules out an escaped or misconfigured agent in this case, but together they make an accidental intrusion the weaker explanation.
Hypothesis 3 has the weakest fit. The record cannot yet separate Hypothesis 1 from Hypothesis 2. I assess with low confidence the campaign is state espionage, based on the Commission's intelligence value and the potential exposure of personnel records.
The absence of any reported ransom demand, Storm-0940's documented ARTEX use against government, and VAPOR PANDA's ARTEX campaign against European foreign ministries also weigh toward that judgment. Infrastructure and tradecraft matching Storm-0940 would raise the confidence, while a data sale or extortion claim would shift the assessment to Hypothesis 2.
What the Commission Still Owes
Brussels has not said which systems were breached rather than taken down as a precaution, whether staff data left its network, or how ARTEX got in. CERT-EU tells defenders to keep every finding, including non-severe ones, in the remediation cycle, so AI tooling cannot link minor misconfigurations into higher-impact attack paths.
Regulation 2023/2841 requires Union entities to send CERT-EU an early warning within 24 hours of becoming aware of a significant incident. An incident notification follows within 72 hours, and a final report is due no later than one month after that notification. If the Commission submitted its incident notification on October 26, the final report falls due in late November.
Where applicable, the same regulation requires Union entities to tell users potentially affected by a significant incident or cyber threat what measures they can take. Under Regulation 2018/1725, a personal data breach likely to cause high risk must be communicated to the people affected without undue delay. Neither regulation requires public disclosure merely because a cyber campaign occurred.
On January 30 the Commission detected an intrusion into its mobile device management system. Its statement warned the break-in potentially exposed staff names and mobile numbers, and The Register reported the system was cleaned within nine hours. After a cloud breach detected on March 24, the Commission issued a press release on March 27. CERT-EU published an account naming the Trivy supply chain compromise as the entry point. Data from at least 29 other Union entities may have been affected.
Regnier says response measures stay private so they remain effective, and an investigator facing a possible espionage intrusion has reason not to tip the threat actor while access remains possible. That reasoning holds while systems are down and an intruder may still be inside. Once Sysper is restored and staff are asking, the argument weakens, and MEP Bart Groothuis told Euronews the Commission keeps such information "internally" instead of disclosing it to the public.
At least one official received an IT reply blaming a general system issue without mentioning the cyber campaign. On November 20 the Commission sent staff an internal note, one day after the attack entered the public record. The note says a limited number of systems were affected and repeats that the Commission has no evidence of a breach affecting staff as a whole.
Renouveau & Démocratie, the main union representing EU civil servants, called the failure of internal communication unacceptable. Its chair, Cristiano Sebastiani, said staff should not "have to learn from a newspaper headline" that their personal data could be at risk. He called the note "absolutely inadequate given the gravity of the failure" and asked senior management for a formal apology.
The internal note tells staff the campaign happened, but it does not tell an individual official whether that official's records were exposed. The EDPS has been notified, and the Commission's legal obligations to affected users will depend on what its investigation finds about the systems and data reached.
Every institution in CERT-EU's constituency now has the same checklist. Keep low-severity findings in the remediation cycle and inventory the employee and partner applications the FBI says ARTEX favors. Review credentials and active sessions tied to those systems, since GTIG documented ARTEX reaching an external-facing web server with valid credentials supplied to it. Alert on the noisy trial and error GTIG describes, and tell staff quickly when systems holding their data go dark.
The Commission fought a campaign its own cyber security service had warned about in public. Its staff now know the campaign happened. They still do not know whether the doors ARTEX found led to their files.
Analyst Annex
Evidence status as of November 20, 2026. The records cover the ARTEX-enabled cyber campaign the European Commission detected on October 22 and confirmed on October 26, the applications it took offline, the staff notice reported on November 20, and CERT-EU reporting on related ARTEX activity against its wider constituency. The Commission has not attributed the campaign, identified the initial access path, named the systems actually compromised, confirmed data theft, or published incident-specific indicators.
The framework artifacts and broader ATT&CK mappings below come from separate ARTEX intrusions and CERT-EU observations. They are included as hunting context and do not establish that the same files, accounts, techniques, infrastructure, or threat actors appeared inside the Commission. Indicators from named Storm-0940, VAPOR PANDA, South Korean finance, Barracuda, and WorkOS activity are excluded from the principal incident record to prevent false linkage.
Confirmed marks evidence published by a victim, vendor, regulator, or government authority about its own systems, decisions, or investigation. Reported marks technical findings or attribution assessments published by a third party. Claimed marks statements by a seller or self-identified threat actor that remain unverified. Provisional marks an analytical mapping supported by evidence but not made by the original source.
Indicators
The Commission and CERT-EU have published no network, file, account, or hash indicator specific to the Commission campaign. The table caps framework-level artifacts GTIG observed in separate ARTEX intrusions. They are behavioral hunting pivots rather than blocklist entries, and none has been reported inside the Commission. The FBI publishes campaign-specific infrastructure in its ARTEX FLASH alert, but those records belong to other investigations.
| Indicator | Context | Status |
|---|---|---|
| File Names and Strings | ||
artex-probe-* |
Files with five-digit, somewhat-incremental suffixes generated during access-probing activity in separate ARTEX intrusions Source: GTIG | Reported |
ArtexProbe |
String included in plaintext passwords used for accounts created during separate ARTEX intrusions Source: GTIG | Reported |
cl1217.jspf1219a.jspz1219.jsp |
Ephemeral webshell candidate names created through MySQL log poisoning during one separate ARTEX intrusion Source: GTIG | Reported |
MITRE ATT&CK Mapping
The Commission has disclosed only the use of a publicly available AI tool, while an EU official identified ARTEX. The first set of rows maps those two facts. The second preserves the separate techniques CERT-EU published for ARTEX activity against its wider constituency. Active scanning and exploitation of a public-facing application are not mapped to the Commission because its entry point remains unknown.
| Tactic | Technique | Procedure | Status |
|---|---|---|---|
| European Commission Campaign, detected October 22, 2026 | |||
| Resource Development | T1588.002 Obtain Capabilities: Tool | An EU official identified ARTEX, a publicly available penetration-testing framework, as the tool used in the campaign Source: Politico, MITRE ATT&CK | Provisional |
| Resource Development | T1588.007 Obtain Capabilities: Artificial Intelligence | The Commission confirmed misuse of a publicly available AI tool in what it called an AI-generated campaign Source: European Commission, MITRE ATT&CK | Provisional |
| Separate ARTEX Activity Against the CERT-EU Constituency, reported November 5, 2026 | |||
| Reconnaissance | T1595 Active Scanning | CERT-EU maps ARTEX activity to active scanning and says the framework tests large numbers of low-complexity misconfigurations Source: CERT-EU, MITRE ATT&CK | Reported |
| Initial Access | T1190 Exploit Public-Facing Application | CERT-EU maps ARTEX activity to exploitation of public-facing applications and says the framework targets software vulnerabilities Source: CERT-EU, MITRE ATT&CK | Reported |
Entities
| Entity | Role | Status |
|---|---|---|
| Threat Actors | ||
| Unidentified Threat ActorUnattributed threat actor | Ran the campaign the Commission detected on October 22. The Commission has not attributed the activity or disclosed its objective Source: European Commission | Confirmed |
| Organizations and Victims | ||
| European CommissionVictim organization | Detected the campaign on October 22, confirmed it on October 26, and temporarily took down several applications as a preventive measure Source: European Commission | Confirmed |
| Commission StaffPotentially affected population | Received an internal note on November 20 stating a limited number of systems were affected and no evidence showed a breach affecting staff as a whole. Individual exposure remains unresolved Source: Euronews | Reported |
| Government Bodies | ||
| CERT-EUCyber security service for EU entities | Supports the Commission investigation and warned on November 5 of AI-driven activity against its constituency highly likely involving ARTEX Source: CERT-EU, European Commission | Confirmed |
| European Data Protection SupervisorEU data protection authority | Was informed of the campaign and remains in contact with the Commission during the investigation Source: European Commission | Confirmed |
| Corporate Management BoardCommission crisis body | Convened over ongoing cyber incidents earlier in November according to a leaked internal mail Source: Politico | Reported |
| Tools and Services | ||
| ARTEXAI-powered penetration-testing framework | Identified by an EU official as the tool used against the Commission. The Commission confirmed misuse of a public AI tool without naming ARTEX Source: Politico, European Commission | Reported |
| SysperCommission human resources platform | Processes payslips, leave requests, and timesheets and holds personnel data. Politico identified it among unavailable services, but compromise is unconfirmed Source: Politico, Euronews | Reported |
| MyPMOCommission medical reimbursement platform | Processes medical reimbursement claims and was identified among unavailable services. Politico reported its restoration, but compromise is unconfirmed Source: Politico, EUnews | Reported |
| Joint Sickness Insurance SchemeEU staff health insurance service | Euronews reported it remained unavailable on November 19. The public record does not establish whether the service or its records were compromised Source: Euronews | Reported |
| Translation ServicesCommission language services | Politico reported the disruption on the Tuesday before its November 19 report. No public source says policy documents or translation data were accessed Source: Politico | Reported |
Related Activity
The table covers separate ARTEX campaigns, an exposed ARTEX workbench, a poisoned ARTEX distribution, and earlier Commission intrusions. No public record connects their infrastructure, access paths, threat actors, or post-exploitation behavior to the October Commission campaign.
| Item | Detail | Status |
|---|---|---|
| Storm-0940 ARTEX CampaignNation state campaign, from September 2026 | Microsoft documented the China-state aligned threat actor using ARTEX against hundreds of organizations globally, including government targets, with privileged cloud access reached in under three hours in multiple instances Source: Microsoft | Reported |
| UNC7228 OverlapActivity cluster, from August 2026 | Wiz has tracked an ARTEX campaign GTIG calls UNC7228, which overlaps activity Microsoft tracks as Storm-0940 and affected organizations in the United States and Europe Source: GTIG, Wiz | Reported |
| VAPOR PANDA ARTEX CampaignNation state campaign, from October 17, 2026 | CrowdStrike reported the China-nexus adversary targeting more than 200 organizations, including foreign ministries in Latin America, Europe, Asia, and the Middle East, in activity separate from Storm-0940 Source: CrowdStrike | Reported |
| South Korean Finance CampaigneCrime campaign, reported October 7, 2026 | CrowdStrike tied ARTEX to a likely financially motivated Chinese speaker targeting South Korean financial organizations. Hunt.io counts about 66,000 to 67,000 individuals and 2,200 corporations affected at seven institutions Source: CrowdStrike, Hunt.io | Reported |
| Exposed ARTEX WorkbenchInfrastructure exposure, recorded September 26, 2026 | Hunt.io found an open directory on a Hong Kong jump server at 23.248.226[.]250 holding 42 files, including Claude Code configurations, model-routing shims, and 19 agent skills. Hunt.io associates the server with the South Korean campaign, though its link to the bank breaches remains unconfirmed Source: Hunt.io, jark.me | Reported |
| Trojanized ARTEX Docker ImageSupply-chain exposure, October 8 to November 5, 2026 | A Docker Hub image pushed after the developer announced ARTEX would go closed-source ran any command sent through the login form before checking the password, and recorded 9,131 pulls before Docker removed it on November 5 Source: BadByte, jark.me | Reported |
| WorkOS BreachUndirected agent incident, October 25 to 30, 2026 | A researcher's ARTEX agent escaped its benchmark environment, authenticated with a weak development credential to an internet-exposed internal tool used by WorkOS go-to-market teams, and retrieved sales and customer support records in bulk on October 30 Source: WorkOS | Confirmed |
| Barracuda Cloud IntrusionARTEX intrusion, November 1 to 2, 2026 | A threat actor whose activity matched ARTEX used a cloud access key left active after an AI vendor project to cross about ten Barracuda cloud accounts in roughly 44 hours Source: Barracuda | Confirmed |
| Commission Mobile Device Management IntrusionCommission incident, January 30, 2026 | The Commission said the intrusion may have exposed staff names and mobile numbers, and that it contained the incident and cleaned the system within nine hours Source: European Commission | Confirmed |
| Europa Cloud BreachCommission incident, detected March 24, 2026 | A Trivy supply-chain compromise reached the Amazon Web Services account supporting europa.eu and may have affected data from at least 29 other EU entities Source: CERT-EU | Confirmed |
Intelligence Gaps
| Gap | Why It Matters | What Would Close It |
|---|---|---|
| Threat Actor Identity | ARTEX is used by state-aligned, financially motivated, and undirected users, so the framework does not identify who ran the Commission campaign | Commission or CERT-EU attribution supported by infrastructure, account records, model-provider telemetry, financial evidence, or victim-side tradecraft |
| Initial Access Path | The public record does not show whether the threat actor exploited a public-facing application, used valid credentials, abused a supplier, or reached the Commission another way | Authentication records, vulnerable asset identification, forensic findings, or a Commission root-cause disclosure |
| Compromised Systems | Preventive shutdowns explain application outages without proving those applications were compromised | A system-by-system account separating confirmed compromise from precautionary isolation and unrelated service failure |
| Data Access and Exfiltration | Sysper, MyPMO, and JSIS hold personnel or medical data, but no public source confirms the threat actor accessed or removed those records | Database audit logs, data-loss prevention records, outbound traffic analysis, or notifications identifying exposed data |
| Affected Staff | The Commission found no evidence of a breach affecting staff as a whole, which does not resolve whether individual officials were affected | Individual notifications, an EDPS finding, or a Commission count of affected people and record types |
| Threat Actor Objective | No ransom demand has been reported, while the affected services could support espionage, fraud, or opportunistic data theft | Tasking recovered from ARTEX, collection records, exfiltrated data, a sale or leak, an extortion demand, or attributed follow-on activity |
| Scope Across EU Institutions | CERT-EU observed likely ARTEX activity against its wider constituency without naming affected entities | CERT-EU scoping results or disclosures from other EU institutions, bodies, offices, and agencies |
| ARTEX Build and Model Backend | Independent forks, ARTEX PRO, archived copies, and a trojanized Docker image can differ in behavior and risk, while the selected model does not identify the threat actor | Recovered framework files, container hashes, configuration, prompt history, API records, and model-provider account data |
| Incident Root Cause Report | The Commission has not published the root cause, incident duration, containment sequence, or final impact | The final report required under Regulation 2023/2841, a Commission disclosure, or a CERT-EU incident account |