High severity

European Commission confirms AI-driven cyberattack that disrupted HR and payroll systems

The Commission says a publicly available AI tool was used against its systems, while Politico and CERT-EU point to ARTEX; whether staff data was taken is still unknown.

By Mr Elite Published 7 min read

Severity
High
Status
Reported by outlets
Sources
3
Last checked

The short version

  • Commission spokesperson Thomas Regnier said on 19 November that the attack was first detected on 22 October and involved a publicly available AI tool.
  • Politico, as relayed by The Record and Security Affairs, reported that CERT-EU identified ARTEX; the Commission has not named the tool.
  • Payroll, leave, medical reimbursement and translation services were reportedly disrupted for more than 10 days.
  • The Commission said it has found no evidence of a data breach affecting staff broadly, but the investigation is open.

What happened

The European Commission confirmed on 19 November that it was hit by a cyberattack involving a publicly available AI tool, with the first signs detected on 22 October, according to spokesperson Thomas Regnier.

Infosecurity Magazine reported that Regnier said the intrusion was confirmed as a cyberattack on 26 October, which triggered mitigation and an investigation. He called it an "AI-generated cyber campaign", according to The Record, but did not name the tool or the attackers' motive.

Politico first reported the incident. According to The Record and Security Affairs, Politico said CERT-EU, the cybersecurity team for EU institutions, identified ARTEX, an open-source AI hacking platform. Security Affairs stressed that ARTEX is a suspected tool, not a confirmed method, and that no one has been identified as responsible.

The disruption reportedly hit personnel systems for more than 10 days. Security Affairs listed Sysper (the HR platform for payslips, timesheets and leave), MyPMO (medical reimbursements), the JSIS health insurance portal and translation services. It said Sysper crashed briefly again on 18 November, and JSIS was still down on 19 November, citing a Commission employee quoted by Euronews. Staff told Politico that pay still arrived on time.

In its statements, the Commission said it took several applications offline as a protective measure and that most have since been restored. It said it had informed affected individuals, was checking other systems, and was in contact with CERT-EU and the European Data Protection Supervisor (EDPS). Regnier said he knew of no ransom demand and that investigators had found no evidence of a data breach affecting staff broadly.

Asked why the incident was not disclosed sooner, Regnier said publicising defensive measures could weaken them. "There was no hiding whatsoever," he told journalists, according to The Record. Security Affairs reported that some staff complained internal messages said little, and one IT reply about Sysper cited a general technical issue.

Why it happened

How the attackers got in has not been disclosed. Security Affairs said there is no public information on the entry point, and the Commission has not said which weakness was used.

What is stated is the method's general shape. Regnier said AI tools were used to detect vulnerabilities, and the Commission said the attack targeted "individual systems". Politico's attribution to ARTEX comes via CERT-EU and has not been confirmed by the Commission itself.

CERT-EU described ARTEX in a technical note on 5 November, according to Security Affairs. An operator gives it a goal in plain language, and the software splits that into smaller tasks carried out by AI agents using real hacking tools. It has no built-in AI model and connects to an external one the operator chooses.

How this kind of failure works

Tools like ARTEX automate the slow, repetitive parts of an intrusion. Finding a weak spot in a network normally means scanning many systems, reading results and trying fixes one by one. An AI agent can do that loop continuously, and it does not tire.

Picture a locksmith who must test every door in a large office block. Done by hand, it takes days and skill. Now give the locksmith a robot that tries every handle, notes which are loose and reports back. The doors are no weaker than before, but they are found much faster and by people with less training.

That is the point SentinelOne's Shaun Williams made, as quoted by The Record: tools like this let attackers find and exploit flaws in internet-facing devices faster, more cheaply and with less skill. The FBI said some intrusions went from first targeting to privileged access, meaning admin-level control, in under three hours.

The defensive consequence is that the gap between a flaw being present and being exploited shrinks. Unpatched, exposed systems and weak access controls matter more when the attacker's search is automated.

Who is affected

Commission staff are the people most directly affected. Security Affairs said some could not submit timesheets or request leave, and the health insurance portal remained down on 19 November. The Commission said it has notified those concerned.

Whether personal data was taken is not known. Security Affairs said it could include names, salary details, addresses or health information, but investigators have not confirmed access. The Commission says it has seen no evidence of a broad staff data breach, which is not the same as confirming none occurred.

Staff can watch for direct notices from the Commission. Wider organisations are affected indirectly: the FBI warned that ARTEX is used against government agencies and dozens of critical sectors, so any internet-facing system is a potential target.

How to defend against it

Tick each step as you check it. Your progress is saved on this device.

What to learn from this

The concept: Attack speed and the shrinking patch window

The time between a vulnerability existing and someone exploiting it has long been the defender's breathing room. Automation reduces it. If a tool can probe, test and exploit without waiting for a skilled person, a flaw left open for a week is a much larger risk than before.

The same principle applies to credential stuffing, where stolen passwords are tried automatically against many sites, and to mass exploitation of newly published flaws. In each case, speed and scale favour whoever has automated.

Defence therefore shifts toward things that do not depend on being faster than the attacker: fewer exposed services, tighter access and detection that triggers quickly.

When attackers automate discovery, the defender's best response is to leave fewer weaknesses exposed and to spot misuse quickly.

Quick check

Sign in to earn 10 XP for a right answer

Why does AI-driven automation make unpatched, internet-facing systems riskier?

Timeline

  1. January 2026 The Commission discloses an earlier incident involving Ivanti software; Security Affairs dates a mobile-device management incident to 30 January.
  2. March 2026 Security Affairs reports a compromised Trivy update was used to enter an AWS cloud environment, with around 91.7 GB of compressed data stolen.
  3. 22 October 2026 First signs of the latest attack are detected, according to Regnier.
  4. 26 October 2026 The intrusion is confirmed as a cyberattack, Infosecurity Magazine reported.
  5. 5 November 2026 CERT-EU publishes a technical note warning that several threat actors are using ARTEX.
  6. 13 November 2026 The FBI issues a Flash notice on ARTEX use against organisations globally.
  7. 18 November 2026 Commissioner Henna Virkkunen briefs the College of Commissioners; Sysper crashes briefly again, per Security Affairs.
  8. 19 November 2026 The Commission confirms the attack publicly; JSIS reportedly remains down.

What is not known yet

  • How the attackers gained access and which weakness they used.
  • Whether ARTEX was actually used, since the Commission has not named the tool.
  • Who is responsible and what they wanted.
  • Whether any staff data was accessed or copied.
  • Whether the attack is linked to the Commission's earlier incidents this year; officials have not linked them.

Questions people ask

Did the European Commission confirm the attack used ARTEX?

No. The Commission confirmed that a publicly available AI tool was used but has not named it. Politico reported, citing CERT-EU, that ARTEX was identified. Security Affairs described ARTEX as suspected, not confirmed.

Was staff data stolen in the Commission attack?

It is not known. Regnier said investigators had found no evidence of a data breach affecting staff broadly, and the investigation remains open. Security Affairs noted that restoring services does not prove data was not copied.

What is ARTEX?

It is an open-source framework, released in July according to The Record, that uses commercial AI models to automate finding and exploiting vulnerabilities. CERT-EU says an operator sets a goal in plain language and AI agents carry it out with real hacking tools.

Which Commission systems were disrupted?

Security Affairs reported Sysper (HR), MyPMO (medical reimbursement), the JSIS health insurance portal and translation services. Most applications have been restored, the Commission said, though JSIS was reportedly still down on 19 November.

Sources

  • The Record: European Commission confirms cyberattack, says hackers used AI tool
  • Security Affairs: European Commission Cyberattack: ARTEX AI Tool Suspected in Prolonged Disruption
  • Infosecurity Magazine: European Commission Confirms AI-Driven Cyber-Attack Took Systems Offline

Know someone who should see this?

Keep reading