European Commission disconnected several applications after an AI-powered campaign

European Commission disconnected several applications after an AI-powered campaign

The European Commission detected a cyber campaign on October 22 that used the ARTEX artificial intelligence tool and, as a precaution, temporarily disconnected several internal applications. The investigation is ongoing, and the institution has not determined who was behind it or whether the attackers accessed employee data.

The Commission’s spokesperson for digital affairs, Thomas Regnier, confirmed the details at the daily press briefing on November 19. He explained that the institution had taken measures «as a precaution and for prevention», and that it was in contact with CERT-EU, the cybersecurity service for European institutions, and the European Data Protection Supervisor.

Regnier declined to characterize the scale of the incident. He said it was a campaign made up of several incidents and that it was still too early to determine its scope. He also confirmed that the attackers used a publicly available AI tool, but did not identify anyone responsible or explain what they were trying to achieve.

Internal services taken offline to stop the spread

The Commission disconnected several applications from its internal systems as a precaution. According to Regnier, most have now been restored. The affected platforms included Sysper, the human resources system used to check salary information, request holidays or leave, submit timesheets, and sign up for training courses.

Staff could not access Sysper for several days. The platform was working again on Thursday, November 19, according to two EU officials cited by Euronews. The Joint Sickness Insurance Scheme (JSIS) was still unavailable that day.

The disruption also affected tools related to viewing payslips, managing working hours, and submitting leave requests, according to information published by TNW. The Commission has not publicly specified which applications were affected or whether the operational problems had other consequences. Euronews described the disruptions as limited for most officials, who would barely have noticed that the systems were affected.

The Commission’s IT department, DG DIGIT, is still examining the incident and trying to determine the damage. The attack was not described as a ransomware incident: Regnier said he was not aware of any ransom demand. The Commission has also not confirmed that the attackers managed to extract information.

Investigation has yet to confirm data access

The main unanswered question is whether the attackers accessed information stored in Sysper, JSIS, or other personnel systems. These platforms may contain names, salary grades, and home addresses, as well as contact details. Euronews noted that these records could facilitate targeted campaigns against employees, but the Commission has not confirmed that they were accessed or stolen.

At the press briefing, Regnier clarified that the institution «has not detected any evidence of a data breach affecting the Commission’s staff as a whole». That response does not resolve whether specific individuals’ data was accessed: the investigation is ongoing, and the spokesperson offered no conclusion on possible access.

«We have not detected any evidence of a data breach affecting the Commission’s staff as a whole».

— Thomas Regnier, European Commission spokesperson for digital affairs

The Commission confirmed the dates at the November 19 briefing: it detected the incident on October 22 and confirmed it on the 26th, following an initial investigation. Regnier said the formal investigation was opened then and that precautionary measures were taken. However, he did not confirm whether the malicious activity was still ongoing. He said several elements had been contained, but avoided declaring the case closed while the analysis continues.

The institution has also not identified the perpetrators. When asked about attribution, Regnier replied that he would not comment because the investigation was open. Asked about the objectives, he said he could not speak on behalf of the threat actor. He also declined to reveal whether the Commission had used AI tools to respond, citing security reasons.

ARTEX automates testing and reconnaissance

The tool used, ARTEX, is an autonomous penetration testing platform powered by language models. CERT-EU explains that it receives a security objective written in natural language, breaks it down into tasks, and allows one or more agents to carry out reconnaissance and assessment steps within a scope defined by the operator.

ARTEX does not include its own AI model: it connects to external language model providers, and each operator supplies a compatible one. The platform can work with real tools and has a dashboard for organizing tasks, logging traffic, reviewing findings, and approving actions. However, CERT-EU warns that it relies on configurable rules and the model to respect the defined scope: there is no technical safeguard that guarantees this.

The European cybersecurity service says it has observed threat actors from different parts of the world using ARTEX to automate campaigns. It also detected similar activity targeting the entities it serves, which it considers very likely to be related to the tool. Its assessment is that ARTEX’s public availability may accelerate its adoption and enable more frequent, faster, and more scalable attacks.

Automation makes it possible to review numerous weak configurations and maintain long attack chains without the fatigue that limits manual testing. CERT-EU notes that it can also facilitate the creation of tailored payloads, including in cases that might not receive as much attention during a manual assessment because of the time involved.

Following several reports of malicious use, the original author took the repository offline in October 2026. The available copies are now independent forks, not an official version maintained by the original project. CERT-EU had warned in an April 2026 publication that autonomous offensive capabilities would gradually reach more actors, and presents ARTEX as an example of this trend.

A new intrusion following the year’s incidents

The incident adds to other incidents affecting the Commission’s systems in 2026. In March, attackers used a stolen API key to access the Amazon Web Services account hosting the Europa.eu website. TNW reported that CERT-EU estimated the stolen information at around 92 GB of compressed data, including names, email addresses, and message content. Euronews put the theft at 91.7 GB and said it included personal information and email content.

The figures published about that theft do not match exactly: TNW puts it at around 92 GB, while Euronews reports 91.7 GB. CERT-EU also said that the March incident may have exposed data from at least 29 EU institutions and bodies.

Another attack, which occurred during the winter, affected the central infrastructure used to manage the Commission’s mobile phones and may have exposed employees’ names and phone numbers. Euronews dates that incident to January. In this context, the new campaign once again puts systems containing personal information under scrutiny, although these services do not normally host the Commission’s internal documents, which are stored on a separate platform, according to Euronews.

Criticism over lack of public communication

Dutch MEP Bart Groothuis, of the Renew group, criticized the Commission for not informing the public about the incident sooner. «The problem I have with the European Commission is that they keep information about this internally and do not make it public», he told Euronews.

Groothuis argued that reporting intrusions at EU institutions, bodies, and agencies would make it possible to explain the bloc’s economic security policy toward its adversaries. At the press briefing, the Commission responded that it had followed its usual procedures: it informed CERT-EU and the European Data Protection Supervisor, as well as notifying the affected services within the institution.

Regnier explained that making the measures public while responding to a campaign could reduce their effectiveness. The investigation remains open, and the Commission continues to take action based on its findings.

The Commission presented a package of measures to strengthen cybersecurity in the EU in January. Virkkunen briefed the other commissioners on the incident on Wednesday, according to TNW.

The campaign’s scope, whether personal data was accessed, and who carried it out remain unclear. The Commission has not confirmed that its employees’ information was stolen and is continuing its investigation.

Sources: TNW | Eu, euronews, Audiovisual Service, CERT-EU