The European Commission, the executive arm of the European Union, has confirmed a cyberattack built on ARTEX, a publicly available AI tool that can hunt for weak spots in internet-facing systems and exploit them largely on its own. For several days, Commission staff were locked out of internal HR and health-insurance platforms. The Commission says it received no ransom demand, and it is still assessing whether any staff personal data was stolen.
What happened
According to Euronews, the Commission's spokesperson for digital, Thomas Regnier, told reporters on November 19 that the attack was based on a publicly available AI tool, named ARTEX, and that it was first detected on October 22. He said the investigation is ongoing, that the Commission has "not detected any evidence of a data breach affecting Commission staff as a whole", and he declined to attribute the attack to any specific actor.
As a precaution, the Commission took several internal apps offline to stop the attack from spreading. Among them was Sysper, the HR platform staff use to check salary information, request leave and submit timesheets. Euronews reports, citing two EU officials, that Sysper was working again on Thursday while the Joint Sickness Insurance Scheme (JSIS) platform remained unavailable. CERT-EU, the cybersecurity service for EU bodies, and the European Data Protection Supervisor have been informed.
Why it matters
The open question is whether the attacker reached the data held in these HR systems: names, pay grades, home addresses and contact details of EU officials. Euronews notes that such a database could fuel far more targeted phishing and social-engineering campaigns against individual staff. Internal documents are hosted on a separate system.
What ARTEX is
CERT-EU published a warning about ARTEX on November 5. It describes the tool as an open-source, LLM-driven autonomous penetration-testing platform: an operator states a goal in plain language and AI agents carry out reconnaissance and testing steps using real tooling. It was built for legitimate security testing, but CERT-EU says it has seen several threat actors worldwide use it to automate campaigns, and that it quickly finds and exploits misconfigurations as well as classic software vulnerabilities. The original author pulled the repository in October 2026 after reports of misuse; the versions in circulation are independent forks that plug into external AI models.
CERT-EU's assessment is that public availability will likely bring more frequent, faster and more scalable automated attacks. The tool's strength is persistence: it can test huge numbers of low-complexity mistakes a human tester would not have time to check.
What to do
- Inventory internet-exposed assets and expose only what truly needs to be public.
- Keep management and admin panels off the internet or untrusted networks.
- Patch internet-facing systems first, and enforce multi-factor authentication on critical services.
- Monitor privileged accounts and alert on bursts of automated scanning in your logs.
For builders, the takeaway is simple: when attackers can run an AI agent against every exposed endpoint around the clock, "nobody will find that forgotten panel" is no longer a defense.

