# European Commission confirms cyberattack that misused a publicly available AI tool

The Commission says an AI-generated campaign detected on 22 October forced several internal applications offline. CERT-EU has linked similar activity to ARTEX, an open-source autonomous pentest framework.

Source: https://junglewise.ai/news/european-commission-confirms-cyberattack-that-misused-a-publicly-available-ai  
Byline: Junglewise News Desk, https://junglewise.ai/news  
Topic: Threat campaigns  
Published: 22 November 2026  
Reading time: 5 min

![The Berlaymont building, headquarters of the European Commission, in Brussels under a clear blue sky](https://hmbbfviplsoqmhdindih.supabase.co/storage/v1/object/public/news-pictures/articles/94f5c0e2-cce7-4a1d-8860-bf33d628a53f/57e14c71-eb80-4a40-b885-bef7c5c89119-1280.jpg)

Picture: [EmDee](https://commons.wikimedia.org/wiki/User:EmDee), [CC BY-SA 4.0](https://creativecommons.org/licenses/by-sa/4.0) [via Wikimedia Commons](https://commons.wikimedia.org/wiki/File:Belgique_-_Bruxelles_-_Schuman_-_Berlaymont_-_01.jpg)

The European Commission has confirmed that it was the target of a cyber campaign carried out with a publicly available artificial intelligence tool, and that it pulled several of its own applications offline in response. Spokesperson Thomas Regnier [made the confirmation](https://audiovisual.ec.europa.eu/en/media/video/I-298222) at the midday press briefing in Brussels on 19 November 2026, saying that "following our investigation, we can confirm that this was an AI generated cyber campaign" and that the tool used was publicly available. Most of the applications taken down have since been restored.

The confirmation came about two weeks after the Commission's own security service, CERT-EU, warned about automated, AI-driven intrusion tooling. The incident was detected on 22 October and confirmed after initial investigation on 26 October. Regnier declined to name the tool, to say which Commission services were targeted, or to attribute the campaign, citing the ongoing investigation. He said that at this stage the Commission has found no evidence of a data breach affecting its staff as a whole, and, asked about a ransom, said he was not aware of any ransomware demand having been made.

[Politico was first to report](https://helpnetsecurity.com/2026/11/20/european-commission-ai-generated-cyberattack-publicly-available-tool) the attack, in an exclusive published the same day as the briefing. According to Politico's sources, Commission defenders detected attackers using ARTEX, an open-source framework for finding weaknesses in IT systems. The Commission has not confirmed that. Politico also reported that the campaign caused more than ten days of disruption and potentially exposed large amounts of data, which Regnier would not endorse: asked whether the attack was massive, he said it was difficult to judge the exact scale while an investigation is still running, and described it only as a campaign made up of a series of incidents.

The practical effects landed on staff. [The Register reported](https://theregister.com/cyber-crime/2026/11/19/european-commission-scrambles-to-contain-ai-assisted-cyberattack/5311103) that Sysper, the human resources platform used for payslips, timesheets and leave requests, and MyPMO, the medical reimbursement system, were taken offline, restored earlier in the week, and that Sysper went down again on the Wednesday. Translation services were also affected. Salaries were paid on time, according to Politico, though some staff could not book time off or submit hours, and whether payslip data was accessed has not been established.

## What to do now

CERT-EU, the cybersecurity service for EU institutions, bodies, offices and agencies, published [guidance on ARTEX](https://cert.europa.eu/blog/from-penetration-testing-to-cyber-threat-beware-of-artex) on 5 November. Its recommendations:

1. Cut the internet exposure of assets that do not need it, and isolate management panels and administrative interfaces so they are not reachable from the open internet.
2. Prioritise emergency patching of public-facing applications.
3. Monitor privileged account activity for unusual use, and deploy endpoint detection and response (EDR) tooling to spot abnormal behaviour on hosts.
4. Implement multi-factor authentication across critical systems if you have not already.
5. Hunt for ARTEX-related indicators in your logs, including automated scanning patterns and tool signatures.
6. On a continuing basis, sanitise error responses so they leak less about your stack, deploy a web application firewall, keep logging and detection practices current, and do not discard minor findings. CERT-EU specifically warns that AI-enabled attackers chain small misconfigurations that a human tester might have filed and forgotten.

## Technical details

ARTEX is an open-source, large language model driven autonomous penetration testing platform written in Go. An operator gives it a security objective in plain language and agent workers then carry out reconnaissance and testing within a defined scope using real tooling. It ships no model of its own and connects to external providers supplied by the operator. According to [heise online](https://heise.de/en/news/Cyberattack-hits-European-Commission-causing-days-of-outages-11531713.html), those include Claude, OpenAI's GPT and Chinese models such as DeepSeek, with the FBI saying attackers most often reach for the Chinese models because they are cheap and appear to lack safeguards.

Scope limits in ARTEX rest on configurable rules and on the model's own compliance. There is no hard technical enforcement.

The original developer, known as Autumn-27, removed the repository from GitHub in October 2026 after reports of misuse. Independent forks and a commercial version continue to circulate. CERT-EU mapped associated techniques to Active Scanning (`T1595`), Obtain Capabilities: Tool (`T1588.002`) and Exploit Public-Facing Application (`T1190`), and said it had already observed comparable AI-driven activity aimed at the EU entities under its protection, most probably carried out with ARTEX. Its assessment: "ARTEX's public availability will likely accelerate adoption by threat actors."

The FBI's Internet Crime Complaint Center published a threat overview of the framework on 13 November. The FBI says actors have abused ARTEX since at least August 2026 to automate reconnaissance, exploitation of misconfigurations and data exfiltration, that researchers found more than 1,600 ARTEX consoles exposed to the internet, and that in some attacks the gap between identifying a target and obtaining privileged access was under three hours. [SecurityWeek reported](https://securityweek.com/european-commission-confirms-ai-assisted-cyberattack-disrupted-services) that the tool has also been used against Barracuda, Railway and ConnectWise customers, among others.

Horizon3.ai chief executive Snehal Antani told The Register that "frameworks like ARTEX let a single actor stand up fleets of agents that plan, probe, exploit, and move laterally with almost no further human input", and noted that shared tooling between state-backed and criminal operators makes attribution harder.

Timeline: detection on 22 October, confirmation on 26 October, an emergency meeting of the Commission's Corporate Management Board in early November, the CERT-EU advisory on 5 November, the FBI alert on 13 November, and public confirmation on 19 November. Regnier said the Commission informed CERT-EU, the European Data Protection Supervisor and affected internal services as usual, adding: "There was no hiding whatsoever. It's just the Commission taken and following the usual procedural steps."

Not yet known: who was behind the campaign, which services were targeted, whether the attack is fully over, whether any staff data was taken, and whether other EU institutions were affected. Regnier said he speaks only for the Commission.

## The story so far

- [Attackers are running AI agent frameworks that turn patches into exploits](https://junglewise.ai/news/attackers-are-running-ai-agent-frameworks-that-turn-patches-into-exploits)

## Related articles

- [CISA adds critical VMware vCenter bypass to its exploited vulnerabilities catalog](https://junglewise.ai/news/broadcom-patches-critical-vcenter-bypass-that-mandiant-says-a-china-linked-group)
- [Russian group Star Blizzard used one hacked US firm to break into another](https://junglewise.ai/news/russian-group-star-blizzard-used-one-hacked-us-firm-to-break-into-another)
- [Node2Shell: CISA names Siemens S7 web API client library among products hit by Brotli flaw](https://junglewise.ai/news/node2shell-siemens-names-scada-and-building-management-products-affected-by)
- [Chrome 158 fixes 204 security flaws, six of them critical](https://junglewise.ai/news/chrome-158-fixes-204-security-flaws-six-of-them-critical)
