European Commission Hit by AI-Assisted ARTEX Attack
Hackers using ARTEX, an open-source AI penetration-testing tool, got into the European Commission’s internal IT, knocking its Sysper HR system and MyPMO medical reimbursement portal offline for more than ten days. The Commission is still working out whether staff data was taken, weeks after CERT-EU warned about ARTEX-driven attacks.
The European Commission is dealing with a cyberattack in which the intruders used an AI agent to do much of the work. According to The Register, which followed a first report by Politico, the Commission’s own security team detected hackers using ARTEX, a publicly available AI-powered penetration testing framework, against its internal IT systems. The disruption has lasted more than ten days. Sysper, which staff use for payslips, timesheets and leave requests, and MyPMO, which handles medical reimbursements, were taken offline, and the Commission’s translation services were also affected. Sysper came back early in the week of November 16, then failed again on Wednesday, November 18.
Politico reports that salaries were paid as usual, though some staff could not book time off or file timesheets. Whether the attackers saw payslips or other personal data is something the Commission has yet to establish. Spokesperson Thomas Regnier told Politico that the Commission “has taken the necessary steps in response to a detected cyber campaign,” that individual systems were targeted, and that, in line with standard practice, those concerned had been informed. Commissioner Henna Virkkunen briefed her colleagues on November 18 after an emergency meeting of the Commission’s Corporate Management Board earlier in the month. The Commission had not answered The Register’s questions about whether data was taken.
What ARTEX is
The Commission’s own cyber agency had warned about the tool. On November 5, CERT-EU, which defends the EU’s institutions, described ARTEX as “an open-source, LLM-driven autonomous penetration testing platform” that takes an objective in plain language and sends out agent workers to achieve it, with no hard technical limit on what they target. It does not ship with its own AI model; operators plug one in. CERT-EU said it had seen similar AI-driven activity against the organisations it protects, which “highly likely” involved ARTEX, and that the tool’s public availability would likely speed up its adoption by attackers. Its advisory does not name the Commission.
Google Threat Intelligence published the most detailed account on October 30. It describes ARTEX as a multi-agent pentesting and command-and-control platform written in Go, with a web console, that splits an attack across six specialised sub-agents and can run Active Directory reconnaissance, escalate privileges, harvest cloud storage keys and build custom webshells. Its developer took the original GitHub repository down in October because of widespread abuse, but forks keep spreading, including one called ScopeWeaver and a commercial version, ARTEX PRO. Google has seen extortion gangs and suspected espionage groups use it since late August, including a campaign that overlaps with the China-linked group Microsoft tracks as Storm-0940. In one intrusion tracked by Wiz, the attackers went from a first compromised account to deep access in about four hours, and in one case in 15 minutes.
Hours, not weeks
The Register also cites an FBI advisory that found more than 1,600 ARTEX consoles exposed on the internet and attacks that went from picking a target to privileged access in under three hours. Snehal Antani, chief executive of Horizon3.ai, told the publication that frameworks like ARTEX mean attacks now happen in hours rather than days or weeks. CERT-EU’s advice is to cut internet exposure, isolate management panels, patch urgently, require multi-factor authentication, watch privileged accounts and hunt for ARTEX artefacts; Google adds short-lived credentials, least privilege, network segmentation and alerts on sudden bursts of activity, since ARTEX fires commands in seconds where a human operator would take minutes.
The case joins a run of incidents in which AI agents did the attacker’s legwork, from an AI-built zero-day that Google caught in May to the rogue agents that reached Indian companies’ data in October. The Commission also enforces the EU AI Act, which puts its own staff among the first to learn how quickly an off-the-shelf agent can work through a large organisation’s systems once it is pointed at them.
Want AI news before everyone else?
The morning's most important AI stories, straight to your inbox. No fluff.