European Commission Confirms 'AI-Generated' Cyberattack: CERT-EU Finds ARTEX AI Pentest Tool Behind 10+ Days of Staff-System Outages, Data Exposure Still Unclear

An intrusion detected on 22 October knocked out the Commission's HR, payroll, medical-reimbursement and translation services; CERT-EU linked it to the open-source ARTEX AI framework, with no attribution yet.
Overview
The European Commission has confirmed it was hit by a cyberattack that its spokesperson Thomas Regnier described as "an AI-generated cybercampaign." The first signs were detected on 22 October 2026. Personnel services were disrupted for more than 10 days, and the Commission took several internal applications offline as a precaution. Most are now back.
According to reporting on the EU's internal investigation, CERT-EU, the EU institutions' cybersecurity service, found that the attackers used ARTEX. ARTEX is a publicly available, open-source AI penetration-testing framework that uses commercial AI models to automate finding and exploiting vulnerabilities. The Commission itself has not named the tool, and it has not named the attackers.
This is the Commission's third disclosed cyber incident of 2026. In January, its central mobile-device management system was breached. In March, the cloud infrastructure hosting the Europa websites was compromised. Officials have not linked the three incidents.
What Was Hit
The attack targeted the Commission's internal HR and staff-service systems:
- Sysper, the HR platform for payslips, working hours, leave requests and training sign-ups. It was unavailable for several days, was restored, then went down briefly again on 18 November.
- MyPMO / JSIS, the staff health insurance and medical-reimbursement services. JSIS was still unavailable on 19 November.
- Translation services, disrupted on 17 November.
These HR platforms hold staff names, pay grades, home addresses and contact details. Internal policy documents are kept on a separate system that would not normally be reachable through them.
Timeline
- 22 October 2026: first signs of the intrusion detected
- Late October to November: more than 10 days of disruption to staff services; applications taken offline as a protective measure
- Early November: CERT-EU identifies the use of ARTEX; the Commission's top internal crisis-management body holds an urgent meeting
- 17 November: translation services disrupted
- 18 November: Sysper briefly down again; Tech Sovereignty and Security Commissioner Henna Virkkunen briefs fellow commissioners
- 19 November: the Commission publicly confirms the attack
Tradecraft
How the attackers got in has not been disclosed, and neither has the vulnerability or system they used. Investigators have not yet established whether ARTEX was used to break in or only for reconnaissance.
ARTEX was released in July 2026 and has quickly become a tool of choice for both criminal and state-backed operators. Its agents chain reconnaissance, vulnerability discovery, credential attacks and exploitation with little human input. The FBI has warned that some ARTEX intrusions went from first targeting to privileged access in under three hours. Reef has tracked several ARTEX campaigns:
- the China-linked VAPOR PANDA operation against 200+ organisations, including European government ministries (intel#130)
- the Storm-0940 campaign against Microsoft 365 tenants
- confirmed breaches at Chainstack and WorkOS (intel#132)
No one has publicly linked the Commission intrusion to any of these campaigns.
Data Exposure
This is the open question. Regnier said: "We have not detected any evidence of a data breach affecting Commission staff as a whole." He also said individual systems were targeted and the people concerned have been informed. Other reporting, citing EU officials, says the attack may have compromised large amounts of personal data. Staff have been given limited details partly because that is still being assessed.
The Commission says:
- it is coordinating with CERT-EU and the European Data Protection Supervisor
- it is checking other systems as a precaution
- its digital services department (DG DIGIT) is investigating
- it knows of no ransom demand
- "The investigation is still ongoing."
On why the attack was not disclosed earlier, Regnier said publicising defensive measures could undermine them: "There was no hiding whatsoever." At least one Member of the European Parliament has criticised the Commission for keeping information about the incident internal.
IOCs
No indicators of compromise have been published for this incident. Reef will update this item if CERT-EU or the Commission releases any.
Detection
For organisations worried about ARTEX-style automated intrusions:
- Watch for bursts of high-rate scanning, password spraying and credential stuffing against internet-facing services, especially from cloud and VPS address ranges. AI agents work at machine speed.
- Alert on the same source probing many unrelated paths or products within minutes. That pattern suggests automated vulnerability discovery rather than a human operator.
- Look for fast escalation: a new session that reaches admin or privileged APIs within hours of first contact.
- Monitor HR, payroll and identity platforms for bulk exports and unusual queries. They hold exactly the personal data attackers want for follow-on phishing.
Recommended Actions
- Patch internet-facing systems quickly. Tools like ARTEX turn newly disclosed vulnerabilities into working attacks within days.
- Put HR, payroll and other staff-data platforms behind phishing-resistant MFA and conditional access, and take them off the open internet where you can.
- Rate-limit and lock out automated login attempts, and block known VPS and anonymising ranges on admin portals.
- Staff of EU institutions should be alert to phishing that uses HR themes such as payslips, leave or medical reimbursements, which could exploit the disruption or any exposed personal data.
- Run a tabletop exercise on an AI-assisted intrusion: assume the attacker reaches privileged access in under three hours, and check whether your detection and response can keep up.
Analyst Assessment
Severity: High. The confirmed impact is operational: more than 10 days without HR and staff services at the EU's executive body. Whether personal data was taken is still unresolved. The bigger signal is strategic. A publicly available AI pentest framework has now been tied to an intrusion at one of the world's most important government bodies, which confirms that agentic AI tooling is moving from security research into routine offensive operations. No attribution has been made. This item will be updated if the Commission or CERT-EU confirms a data breach, an attacker or indicators.
