On November 19, 2026, Thomas Regnier, spokesperson for the European Commission, confirmed at a regular press briefing in Brussels that the Commission had been hit by a cyberattack carried out with the help of publicly available AI tools. Several internal systems were taken offline, and some services have been disrupted for more than ten days. The incident itself is small, but it puts a long-debated question on the table: publicly sold AI security tools are becoming off-the-shelf weapons for attackers.
What the Commission confirmed — and what it did not
According to Regnier, the attack was detected on October 22 and confirmed as a cyberattack on October 26, after which an investigation was opened and protective measures were taken. The systems affected were Sysper, the HR platform for payslips, timesheets and leave, and MyPMO, which handles medical reimbursements; translation services were also hit. Most systems have since been restored. Regnier also drew a clear line: so far there is no evidence of a data breach affecting staff as a whole, no ransom demand has been received, the European Data Protection Supervisor (EDPS) has been notified, and the investigation continues.
What ARTEX is, and why it is being named
Politico, which first reported the story, cited EU officials as saying the attackers used ARTEX, a publicly available AI penetration-testing framework. The distinction matters: the Commission's spokesperson only confirmed that an AI tool had been misused and did not publicly name ARTEX; the link between the two comes from monitoring by CERT-EU, the EU institutions' cybersecurity team, and from officials' accounts. ARTEX itself ships with no model: written in Go with a web console, it lets an operator set objectives in plain language, after which AI agents break the task down and call real hacking tools to carry it out. AI agents finding vulnerabilities is nothing new — Google previously reported agents completing vulnerability discovery and exploitation fully autonomously; this year also saw an AI find a serious Java vulnerability within an hour, sending servers worldwide scrambling to patch overnight. As early as November 5, CERT-EU warned that tools of this kind were being used against EU institutions; the developer pulled the project from GitHub in October, but forks are still circulating.
The incident should be judged on two levels. On the facts: a nearly month-long intrusion that relied not on a secret weapon but on tools anyone can download — which means defenders can no longer assume attackers face a high barrier to entry. On the institutional level: the Commission's own crisis management board held an emergency meeting, and tech commissioner Henna Virkkunen briefed the full College of Commissioners on November 18 — when AI tools serve both attack and defense, an institution's incident response, permission controls and staff data protection all have to be rewritten for the new threat model. The investigation is not over; who the attackers were and whether any data was stolen remain unknown.