For more than ten days, thousands of European Commission staff could not do something utterly ordinary: check their payslips. The human-resources portal was down. Leave requests went nowhere. Medical reimbursements sat in limbo, untouched.
Behind those blank screens, the European Union’s executive arm was fighting a cyberattack. And the weapon, investigators believe, was an AI tool built to break into computer systems.
The European Commission cyberattack came to light on Thursday, November 19, when Politico reported that hackers had disrupted the Commission’s internal systems for over ten days. Hours later, the Commission confirmed at its midday press briefing in Brussels that it had responded to a cyber campaign carried out with an AI tool.
According to Reuters, it was the first time the EU had publicly reported a cyberattack that used an AI tool.
What the European Commission cyberattack hit
The disruption struck the unglamorous machinery that keeps a huge institution running. Sysper, the Commission’s HR platform, is where staff view their payslips, record their working hours and request leave. MyPMO handles medical reimbursements. Both were taken offline as a precaution.
Translation services were disrupted too, on November 17. Most of the affected applications were brought back online earlier in the week. Then Sysper crashed briefly again on Wednesday, November 18, a reminder that recovery from an incident like this is rarely a straight line.
None of this sounds dramatic. There were no flashing screens, no ransom notes, no leaked documents splashed across the internet. That is precisely what makes the incident unsettling. The systems that stopped working were the quiet, everyday ones — and nobody outside the Commission noticed for more than a week.
ARTEX: the AI tool at the centre of the investigation
CERT-EU, the cybersecurity team that protects EU institutions, detected the activity and identified ARTEX as a possible tool used in the attack. ARTEX is an open-source, LLM-driven penetration-testing framework.
It does not include its own AI model. Instead, it connects to an external AI model chosen by the operator, which coordinates agents that probe a target’s systems for weaknesses.
Penetration-testing tools are meant to help defenders find holes in their own networks before criminals do. But the same automation that makes them useful to defenders makes them dangerous in the wrong hands.
An AI agent does not get tired, does not sleep, and can test thousands of attack paths in the time a human tester would spend on one.
One thing needs to be said plainly. ARTEX is a suspected tool, not a confirmed attack method. No public information explains how the attackers first entered the Commission’s systems, and no threat group has been named.
Its developer pulled the project’s GitHub repository in October, after reports that hackers were putting the software to use. Independent forks remain publicly accessible, so taking down the original did not remove the tool from circulation.
The warnings that came before the attack
The Commission’s own cybersecurity team had been warning about this exact threat. In a technical note published on November 5, CERT-EU described ARTEX as a system that uses AI agents, and said it had spotted what it believed to be ARTEX-assisted attacks against organisations under its watch.
The campaign was detected on October 22 and confirmed after an initial investigation on October 26. The warnings and the attack were unfolding at almost the same time.
The FBI sounded a similar alarm days later. In an advisory issued on November 13, it revealed that researchers had found more than 1,600 ARTEX consoles exposed to the internet. Some campaigns using the tool had progressed from initial targeting to privileged access in under three hours.
Wasl Tech covered the FBI’s ARTEX warning earlier this month.
Three hours from first probe to full access. That is the speed AI brings to an attack, and it is why the timeline of this incident matters. Traditional defences assume attackers need time. When the attacker is an agent, that assumption breaks.
The question nobody can answer yet: was staff data stolen?
The biggest unanswered question is whether staff data was stolen. The systems that were disrupted hold exactly the kind of information attackers prize: names, salary details, home addresses and health information. A breach of the medical reimbursement system would be especially sensitive.
So far, the Commission has not confirmed any data theft. At the briefing, spokesperson Thomas Regnier said that at this stage no evidence had been detected of a data breach affecting Commission staff “as a whole”.
That qualifier matters. It is not a statement that no personal data was exposed. It is a statement about what investigators have found so far.
The Commission says it is in close contact with CERT-EU and the European Data Protection Supervisor, and that the supervisor has been informed. If the investigation later shows that staff data was taken, tens of thousands of employees would need to be told what was exposed.
Crisis meetings at the top of the Commission
The incident reached the highest levels of the institution. The Corporate Management Board, the Commission’s top internal crisis body, met urgently over the incident earlier in November, and the digital services department ordered extra security checks.
On November 18, technology commissioner Henna Virkkunen briefed the full College of Commissioners. That is the moment a technical incident becomes a political one.
Regnier confirmed the response to Politico in measured terms: “Individual systems were targeted and, in line with our standard practice, we have informed those concerned.” The Commission has declined to characterise the scale of the campaign, and has said nothing about who might be behind it.
Why the European Commission cyberattack matters
This is not the first time AI has been used in a cyberattack. But it is the first time the European Union has publicly confirmed that one of its own institutions was hit by an AI-assisted campaign. That makes it a landmark, and a warning to every other government and company watching.
The context is a threat landscape that is changing fast. Google’s threat intelligence researchers recently reported the first observed fully automated vulnerability discovery and exploitation, with no human in the loop. Wasl Tech covered Google’s findings on autonomous AI cyberattacks earlier this month.
Microsoft’s researchers, meanwhile, showed that reprogramming fewer than ten dormant neurons in a language model can turn an AI agent rogue, with no retraining and no poisoned data. Read Microsoft’s Lazarus backdoor research to see how small the change needs to be.
And the Fireworks AI breach showed that AI infrastructure itself has become a target, with customer datasets and model weights exposed.
The pattern is consistent. AI is lowering the cost of attacking and raising the cost of defending. A single operator with an AI agent can now do the work that once needed a team. The European Commission, with all its resources, still spent more than ten days restoring its systems.
What happens next
The investigation is ongoing. Most of the affected applications are back online, but the brief relapse of Sysper on November 18 shows how fragile that recovery can be. Until investigators establish how the attackers entered and whether any data left the building, the incident is not truly over.
Several questions will decide how this story ends. Did the attackers get in through an exposed ARTEX console, a stolen credential, or something else entirely? Was any staff data copied before the systems were taken down? And will the public forks of ARTEX be dealt with, or will the tool keep spreading to new operators?
For now, Commission staff can log back in and check their payslips. But the institution still does not know what, if anything, walked out the door with the attackers.
Source: The Register



