ThinkFacility Sign in
  1. Home
  2. News
  3. AI policy

European Commission confirms a cyber campaign run with a public AI tool, detected October 22

Its spokesperson wouldn't name the tool, but CERT-EU had warned on November 5 that attacks on EU bodies highly likely involved ARTEX.

Part of a running story: ARTEX, the hacking agent that keeps turning up (5 parts)

On November 19, 2026, the European Commission confirmed that it had been hit by what its spokesperson called an AI generated cyber campaign. Thomas Regnier said it from the podium at the Commission's midday press briefing in Brussels, after a reporter asked about a Politico story on the attack.

The Commission had kept quiet about it for almost a month. Regnier said the incident was detected on October 22 and confirmed on October 26, and that the investigation is still going.

Detected
22nd of October
Confirmed
26th of October, following the initial investigation
What was used
an AI tool that is publicly available
Staff data
no evidence of a data breach affecting Commission staff as a whole
Who did it
Regnier wouldn't go into attribution

What the Commission said

As a protective step, the Commission took several applications offline for a while, and Regnier said most are back. He wouldn't say which ones. He also wouldn't put a size on it, and when a reporter asked whether "huge" was the right word, he said the scale was hard to judge while the investigation runs.

He did describe it as a campaign, meaning a series of incidents, and he was firm on one point.

we can confirm that this was an AI generated cyber campaign.

From Audiovisual Service

He wasn't aware of any ransom demand. On whether the Commission used AI to defend itself, he declined to say, for security reasons. Asked why staff heard about it from the press, he said the Commission had informed the affected services and the European Data Protection Supervisor, and that "there was no hiding whatsoever."

Where ARTEX comes in

Regnier never named the tool. He called it an AI tool that's publicly available, and that's as far as the Commission's own words go.

The name in the coverage, ARTEX, traces back to CERT-EU, the cybersecurity service for the EU's institutions. On November 5 it published a warning about ARTEX, an open-source penetration-testing platform driven by AI models, and said it had already seen this.

CERT-EU has observed similar instances of AI-driven activities against its constituency. These activities highly likely involved the use of ARTEX.

From CERT-EU

That post doesn't name the Commission, and the briefing doesn't name ARTEX. So we can't tell you from first-hand sources that this attack used it. (Politico reports that CERT-EU found it did. We couldn't find that finding published.)

ARTEX has turned up in our coverage before: in Microsoft's account of the ConnectWise breach, and in CrowdStrike's on South Korean banks. CERT-EU says its original author pulled the repository in October 2026, and that the copies still around are independent forks. It also notes that ARTEX ships with no model of its own. Operators plug in whichever one they like.

What it means outside Brussels

The people most directly affected work at the Commission, and some of their applications went dark for a while. Regnier's answer on their data was narrow. He said there's no evidence of a breach affecting staff "as a whole," which doesn't tell any one person whether their own records were touched.

For everyone else, CERT-EU's post is the useful part, because ARTEX works the same way against a city council or a small business. CERT-EU says the tool is good at checking huge numbers of low-level misconfigurations that a human tester would spot but never have time to check. Its first advice is plain: cut what's exposed to the internet, keep admin panels off it, patch exposed systems first and turn on multi-factor authentication everywhere critical.

The tool relies on configurable rules and the model itself to respect scope and avoid destructive actions; there is no hard technical enforcement.

From CERT-EU

I'd read that as the reason small findings now matter. CERT-EU tells organizations to keep even non-severe findings in the fix list, so AI tooling can't chain minor misconfigurations into a bigger attack.

Regnier said the Commission won't make conclusive remarks while the investigation is open, and he gave no date for when it will close.

More on AI policy

All AI policy stories