European Commission confirms cyberattack assisted by AI

European Commission confirms cyberattack assisted by AI

The European Commission confirmed that an artificial intelligence-assisted cyberattack campaign forced it to temporarily take several internal applications offline. The incident was detected on October 22 and confirmed four days later; the investigation is ongoing, and the institution says it has found no evidence of a breach affecting its staff as a whole.

The confirmation came at the Commission’s daily press briefing on November 19, almost a month after the initial detection. Thomas Regnier, a Commission spokesperson, explained that the institution had acted to contain the campaign and that most of the applications taken offline had already been restored. He did not specify which actor was behind it or define the scale of the attack.

Internal applications disconnected

The affected services include staff management tools, which allow employees to view salary information, request leave and handle other work-related matters, as well as systems related to medical reimbursements and translation. SecurityWeek identifies Sysper, the human resources platform, and MyPMO, the medical reimbursement system. Hardware Upgrade mentions Sysper and JSIS, which is linked to staff health insurance.

The Commission did not provide a complete list of the affected systems at the briefing. Regnier said that several applications had been temporarily disconnected and that most were available again. According to information reported by SecurityWeek and MuySeguridad, disruptions to some services lasted more than ten days.

The spokesperson described the incident as a campaign, meaning a series of incidents, but declined to assess its scale while the investigation remains open. He also did not confirm whether the attackers had managed to access the systems or what their objectives were. The Commission said that some containment measures had already been taken, without declaring the investigation over.

What is known about staff data

So far, the Commission says it has found no evidence of a data breach affecting staff as a whole. That statement does not resolve whether information about specific employees may have been exposed, or to what extent: the institution is keeping the investigation open. Regnier also said he was not aware of the Commission having received a ransom demand.

The uncertainty has worried staff at the institution. Euronews reported that some officials had expressed concern that their data might have been leaked, and cited an anonymous employee who said they feared this possibility, also because of the phishing emails they receive every day.

Renouveau & Démocratie, the European Union civil servants’ main trade union, criticized the fact that staff had not been informed when the attack was detected. Its president, Cristiano Sebastiani, described the internal note sent to employees as «absolutely inadequate» and called on management to issue a formal apology, according to Euronews.

«Nothing was concealed».

— Thomas Regnier, European Commission spokesperson, according to MuySeguridad

The Commission said it had followed its usual procedures and notified CERT-EU, the cybersecurity service for the European Union institutions, and the European Data Protection Supervisor. Regnier explained that disclosing details of defensive measures earlier could have reduced their effectiveness. The institution publicly announced the incident on November 19, after the press had reported on it.

The role attributed to ARTEX

Regnier confirmed that the campaign used a publicly available artificial intelligence tool, but did not name it at the press briefing. ARTEX was identified in press reports and in an earlier CERT-EU alert: the service had observed activity targeting its environment that, in its assessment, probably involved the use of this tool.

CERT-EU describes ARTEX as an open-source penetration testing platform that uses external language models and agents to autonomously carry out reconnaissance and testing tasks within a scope defined by its operator. It does not include its own AI model. The operator configures the model and directs the work from a dashboard; the available public versions are independent forks, after the original author took down the repository in October 2026 following reports of its misuse.

The tool can search for misconfigurations as well as software vulnerabilities, and chain together reconnaissance and exploitation steps. CERT-EU warns that its rules and model guide it to respect the defined scope and avoid destructive actions, but there is no technical barrier to guarantee this. Capabilities may vary between different forks.

The Commission has not clarified whether ARTEX was used to exploit vulnerabilities or only to locate weaknesses during reconnaissance. Nor has it identified those responsible. At the briefing, Regnier confirmed the use of an AI tool but said he would not discuss attribution while the investigation was ongoing.

An earlier warning and other incidents

CERT-EU had warned about ARTEX before the Commission confirmed the campaign. The service said that the tool’s public availability could accelerate its adoption by malicious actors and make automated attacks more frequent, faster and more scalable. It also recalled that it had already warned in April 2026 about the expansion of autonomous offensive capabilities.

MuySeguridad reports that the FBI warned that various actors had used ARTEX since August against financial, government, technology and healthcare organizations. The outlet also attributes to some cases intrusions that progressed from initial contact to privileged access in less than three hours. The Commission has not linked its incident to any of those actors or campaigns.

The attack adds to other incidents that have affected the European Union’s digital infrastructure. Euronews reported that, in March, attackers compromised the cloud infrastructure hosting the Europa.eu websites and stole around 91.7 GB of data. In January, another attack targeting the Commission’s core mobile device infrastructure may have exposed employees’ names and phone numbers.

Hardware Upgrade calls the October incident the third cyberattack the Commission has made public in 2026, following the incidents in January and March. At the briefing, Regnier also declined to speak on behalf of other European Union institutions or confirm whether they might have been affected.

The investigation remains open

The Commission is keeping in contact with CERT-EU and the European Data Protection Supervisor as it examines what happened and whether any information was exposed. Regnier said that further measures would be taken as results came in. The institution has not published any conclusions about the initial access, the data that may have been affected or the attackers’ identity.

Most of the disconnected applications had already been restored when the Commission reported the incident. But the press briefing did not establish whether the campaign had ended: Regnier said that some elements had been contained and that the investigation remained open, so he avoided drawing a definitive conclusion about its scope.