European Commission confirms AI-assisted cyberattack on services
The European Commission confirmed an AI-assisted cyberattack disrupted internal applications for over 10 days in late October. Spokesperson says no evidence of a staff data breach.
The European Commission has confirmed it was hit by an AI-assisted cyberattack that disrupted several internal applications. Spokesperson Thomas Regnier disclosed the incident at a November 20 press conference, saying the Commission "temporarily took down several applications, and most of them have since then been restored" and that the attack is "largely contained." The Commission has not published a written advisory — this confirmation came from the press conference, first surfaced in reporting by SecurityWeek and Politico.
What happened
Per SecurityWeek, the attack hit in late October, was detected on October 22 and confirmed internally on October 26; Politico reported the disruptions ran more than 10 days. Politico identified the affected systems as the Commission's HR platform Sysper, its medical-reimbursement system MyPMO, and its translation systems, and reported that large amounts of data may have been compromised. Regnier said the investigation found no evidence of a data breach affecting Commission staff, and that he was not aware of any ransom demand.
The "AI-assisted" label is the Commission's own: Regnier confirmed the attackers used an AI tool but did not name it or explain how it was used. No threat actor has been attributed.
The Artex angle
About two weeks before the disclosure, CERT-EU warned about threat actors adopting Artex, an AI-powered penetration-testing framework. CERT-EU's assessment: "Artex's public availability will likely accelerate adoption by threat actors." Neither CERT-EU nor the Commission has said Artex was used in this attack, so treat the link as context, not attribution. Artex has already surfaced in intrusions at Barracuda, Railway and ConnectWise customers, among others — see our coverage of the Railway environment-variable breach and of autonomous AI pentest tooling finding zero-days.
Reading the tea leaves
"AI-assisted" is doing a lot of work in the official framing and almost none in the technical record: the Commission named no tool, no technique and no actor. It may mean an attacker used an LLM-driven framework like Artex to accelerate reconnaissance and exploitation, or it may be a convenient label for a messy intrusion. Until the Commission or CERT-EU publishes specifics, the useful signal is the pattern — AI-accelerated offensive tooling is now cited in named-victim incidents, not just vendor demos.
What to do today
- If you run internet-facing or widely integrated internal apps (HR, reimbursement, translation gateways), review segmentation so a single application compromise can't cascade across shared identity and data stores.
- Assume AI-accelerated recon against your perimeter: tighten rate-limiting, anomaly detection on auth endpoints, and alerting on unusual enumeration — the tempo these frameworks enable is the differentiator, not novel exploits.
- Read CERT-EU's Artex advisory and map its noted TTPs to your detections.
- Rehearse the "take the app offline" decision the Commission made: know which services you can pull without halting operations, and who signs off.