European Commission AI Cyberattack: What It Has Confirmed

The European Commission says an 'AI generated cyber campaign' using a publicly available AI tool forced it to take several applications offline after detection on October 22. Here is what the Commission confirmed, what is only reported, and what engineers can check.

Three European Union flags on tall poles in front of a glass-louvred office building
Photo by Carl Gruner on Unsplash

What happened: The European Commission has confirmed that a cyber campaign which misused a publicly available AI tool forced it to take several applications offline. Spokesperson Thomas Regnier told the Commission’s midday press briefing on Thursday, November 19, 2026, that “following our investigation, we can confirm that this was an AI generated cyber campaign,” according to Help Net Security. The incident was detected on October 22 and confirmed on October 26.

Regnier said that “at this stage of the investigation, we have not detected any evidence of a data breach affecting Commission staff”, and that most of the applications taken down “have since then been restored.” Politico first reported the attack; the Commission’s statements were also reported by SecurityWeek. The briefing video is on the Commission’s audiovisual service.

What the Commission said

All quotes below are from Regnier’s remarks as reported by Help Net Security and SecurityWeek on November 20.

  • The attack: “a misuse of an AI tool”, and “it was an AI tool that is publicly available.” He did not name the tool.
  • Scale: “I’m not going to qualify the scale and extent of this cyber attack.” He added: “It is a campaign, so there was a series of incidents indeed.”
  • Services: “As a protective and preventive measure, we temporarily took down several applications, and most of them have since then been restored.” He did not say which services were targeted.
  • Data: no evidence “of a data breach affecting Commission staff” so far.
  • Containment: “many elements have been contained, so that’s the good news.” He would not say whether the attack has ended.
  • Ransom: “I’m really not aware of any ransomware request made to the Commission, so definitely not.”
  • Attribution: “So on the attribution, I will not go there.”
  • Who is involved: the Commission is in “close contact with both CERT-EU and the European Data Protection Supervisor”, and the investigation continues.

Timeline

Date (2026)EventSource
October 22Incident detectedCommission (Regnier)
October 26Incident confirmed after initial investigationCommission (Regnier)
November 19Politico reports the attack; Commission confirms at midday briefingHelp Net Security, SecurityWeek
November 20Coverage of the Commission’s confirmationHelp Net Security, SecurityWeek

Confirmed vs not confirmed

ClaimStatusWho says so
A cyber campaign misused a publicly available AI toolConfirmedCommission
Several applications taken offline; most restoredConfirmedCommission
No evidence so far of a breach of staff dataConfirmed (at this stage)Commission
No ransom demand knownConfirmed (spokesperson “not aware”)Commission
The tool was ARTEXNot confirmedPolitico, as relayed by Help Net Security
Affected services included HR platform Sysper, medical reimbursement system MyPMO and translation systemsNot confirmedPolitico, as relayed by SecurityWeek
Attackers potentially compromised large amounts of data; disruption lasted over 10 daysNot confirmedPolitico, as relayed by SecurityWeek
A specific group or country is responsibleNot stated by anyone on recordCommission declined to attribute

The Commission’s “no evidence” statement and Politico’s report on data are not necessarily in conflict: one is about staff data found so far, the other about what may have been accessed. Expect this to change as the investigation continues.

What ARTEX is

The Commission has not linked the incident to a named tool, but Politico’s report and CERT-EU’s earlier warning have put ARTEX in the frame. Help Net Security says CERT-EU warned on November 5 about “an AI-powered penetration-testing framework”, and SecurityWeek quotes CERT-EU: “We assess that Artex’s public availability will likely accelerate adoption by threat actors, resulting in more frequent, faster, and more scalable automated attacks with a higher success rate.”

What researchers have published about the tool:

  • Dark Reading describes ARTEX as “an open source, multiagent AI penetration testing framework written in Go”, published to GitHub in July by an independent developer, and “by intention, a legitimate tool”. It says the tool “is capable of brute-forcing passwords, dumping databases, enumerating and abusing APIs, and more”, and that the creator deleted the original repository at the end of October “due to rampant abuse, but many forks of the tool remain.”
  • SentinelOne, on November 13, called it “an agentic AI penetration testing platform” that by default drives off-the-shelf tools such as Nmap, curl, wget and headless Chromium. It reports that “actively maintained forks and backups of the original repository remain publicly available”. SentinelOne’s account of when the project went closed source (October 8 in its timeline, “mid-October” in its summary) differs from Dark Reading’s account of the repository deletion.
  • SentinelOne also cites CERT-EU as reporting activity against EU institutions that “highly likely involved the use of ARTEX”. We have not been able to read CERT-EU’s own post.

What it means for engineers

The Commission has not said how the attackers got in, so there is no confirmed technique to defend against here. What the case does show is a defender’s problem that does not depend on the details: an open-source agent that chains reconnaissance, exploitation and credential abuse lowers the effort of trying every exposed weakness at once. Public-facing applications that were “too obscure to target” are no longer protected by that.

Checks drawn from SentinelOne’s published guidance on ARTEX activity, which you can run now:

  • Alert on the tool’s footprint. SentinelOne recommends alerting “on user agents that contain artex”. Expect forks to change this, so treat it as a cheap tripwire, not a control.
  • Take admin interfaces off the internet. “Put administrative interfaces behind a VPN.” An agent that scans everything finds every login page.
  • Rotate secrets you find, and block new ones. SentinelOne advises: “Rotate every secret you find” and “Enable GitHub push protection”. Our guide to detecting and rotating leaked API keys covers the workflow.
  • Watch Entra ID secret failures. SentinelOne lists sign-in errors AADSTS7000215 (invalid secret) and AADSTS7000222 (expired secret), and says: “Treat a failure followed by a success as urgent.”
  • Patch exposed bugs on exploitation signals, not schedules. SentinelOne’s attack chain starts with “IDOR, command injection, and SSTI flaws”. Automated discovery shortens the time between disclosure and attempted exploitation; see how to alert on new CISA exploitation data in CVE records.

What we will update

This page will be updated with a dated note if the Commission names the services affected, revises its data-breach assessment, confirms the tool, or if CERT-EU publishes findings.

Sources

Frequently asked questions

What happened in the European Commission cyberattack?

The Commission says a cyber campaign that misused a publicly available AI tool was detected on October 22, 2026 and confirmed on October 26. As a protective measure it took several applications offline, and most have since been restored. The investigation is ongoing with CERT-EU and the European Data Protection Supervisor.

Was data stolen from the European Commission?

The Commission says that at this stage of the investigation it has not detected any evidence of a data breach affecting its staff as a whole. Politico reported that the attackers potentially compromised large amounts of data. That report has not been confirmed by the Commission.

Which AI tool was used against the European Commission?

The Commission has not named it. Spokesperson Thomas Regnier said only that it was a publicly available AI tool. Politico reported that CERT-EU detected the attackers using ARTEX, an AI penetration-testing framework, but the Commission has not confirmed this.

Who is behind the attack on the European Commission?

No one has been named. Regnier said he would not go into attribution and did not answer a reporter's question about whether the tool was Chinese. As of November 21, 2026, no attribution appears in the Commission's on-record statements.

Hacking NewsEuropean CommissionAI SecurityArtexCert-EuIncident Response