ComputerWeekly

AI red team tool Artex likely behind cyber attack on EU body


Artex, an open-source AI-powered penetration-testing framework, may have been implicated in a cyber incident affecting the European Union’s (EU’s) executive body, the European Commission (EC) in Brussels.

The attack, which unfolded at the end of October, appears to have targeted the Commision’s Sysper human resources (HR) platform, leaving staffers unable to undertake some key tasks. The EC swiftly took down affected systems and Computer Weekly understands Sysper is back online at the time of writing, although according to Euronews other systems – including a health insurance programme – are still offline.

Speaking on Thursday 19 November, Commission spokesperson for digital Thomas Regnier said: “The investigation is still ongoing. We are taking any necessary steps based on the findings as they come…. We have not detected any evidence of a data breach affecting Commission staff as a whole.”

The European Union’s (EU’s) Computer Emergency Response Team (CERT-EU) said: “CERT-EU has observed similar instances of AI-driven activities against its constituency. These activities highly likely involved the use of Artex.”

ISACA chief global strategy officer Chris Dimitriadis said: “On the eve of the EU-UK summit, the news that the European Commission has suffered a significant cyber attack exposes Europe’s executive body. 

 “While alarming, this attack is not surprising. Thanks to AI, attackers now hack at the speed of intent – a pace the public and private sector cannot outrun without serious investment in the cyber workforce. Until organisations deploy AI defensively as fast as criminals deploy it aggressively, we will continue on this trajectory. 

“In the short term, the Commission will investigate whether staff data has been compromised and what weakness hackers were able to use Artex to exploit. They should be bolstering their cyber security and building long-term resilience by investing in recruiting, training, and upskilling privacy professionals who can build defences that are stronger than the AI tools now freely available to attackers.” 

What is Artex?

At its core, the legitimate Artex framework consists of a large language model (LLM)-powered agentic platform, written in Go with a Next.js frontend.

It can be used to automate offensive cyber tasks and was recently identified by threat intel teams and law enforcement agencies as increasingly finding favour among cyber criminals thanks to its ability to shrink the time needed to chain various offensive cyber tasks.

Indeed, some intrusions orchestrated using Artex are believed to have moved from initial targeting to full privileged access in less than three hours, moving up cyber attack timelines.

But its value to a threat actor lies in not just its speed, but its scale and persistence as well. Artex is able to probe huge numbers of low-complexity issues, such as misconfigurations, that human eyes would miss, and of course it does not get tired or need tea breaks.

Microsoft said that the tool was being heavily used by a Chinese state-aligned threat actor tracked as Storm-0940, which has used Artex to obtain access to its victims’ cloud infrastructure – including Microsoft 365 environments. Redmond’s findings have been corroborated by Google’s Threat Intelligence Group (GTIG) and its Mandiant stablemate.

Artex was first released on GitHub by a user known by the handle Autumn-27 in July of this year and reached between 1,500 and 2,000 users in the space of 10 weeks, according to the FBI in the US.

Autumn-27 had allegedly stated that Artex was supposed to be used for “authorised security testing” within defined scopes, and in fact pulled its repository on 8 October after reports it was being abused.

Despite this, Artex persists and continues to proliferate through community forks, localised variants, and a commercial version, Artex Pro.

CERT-EU said: “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.”

“In April 2026, in our blog post AI is changing the economics of vulnerability discovery, we warned that autonomous offensive capabilities would progressively reach a wider range of actors. Artex is a concrete example of that shift materialising.”

Urgent steps for cyber teams

For security teams, CERT-EU has made available a list of tasks to undertake to safeguard against Artex.

In the immediate future, priorities should be to audit publicly exposed IT assets and limit exposure; to ensure management and control panels are isolated from the public internet or untrusted networks; to priorities emergency patching for all assets that must be exposed to the internet; and to monitor privileged and admin account activity for evidence of suspicious behaviour or abuse.

In the short-term, security teams should beef up their use of multifactor authentication (MFA), and hunt for and alert on connections to known or suspected threat infrastructure to identify potential compromises, particularly artefacts and behaviours consistent with Artex. Endpoint detection and response (EDR) tools can also be pressed into service to block strange processes.

On a continuous basis, security teams should sanitise error responses from internet-exposed devices to minimise threat actor enumeration of their assets, protect said assets with web application firewalls (WAFs), adopt detection and logging best practice for incident response, ensure monitoring covers AI tools and assets using them, conduct regular vulnerability scans and penetration tests, and retain all findings – even those that don’t seem very severe – in the remediation cycle, which can help stop tools like Artex from chaining a series of minor problems into a great big one.

What next for Artex?

Looking ahead, Snehal Antani, CEO of Horizon3, said the clear jump in attack volume and velocity seen in recent weeks was likely attributable to frameworks like Artex and similar tools than enable bad actors to harness fleets of agents, essentially industrialising cyber attacks using proven, basic techniques.

“Artex and the forks that followed it are the multiplier. One operator can now run many targets in parallel, and the agents keep iterating even when the human walks away,” said Antani.

In emailed comments, Antani told Computer Weekly that Artex and similar frameworks will drive a type of attack standardisation that makes attribution much harder.

“When different groups – state-aligned or purely criminal – are all running near-identical agent harnesses, models, and playbooks, their operational fingerprints start to look the same. You lose a lot of the traditional signals that used to separate actors.

“Going forward, expect this to become normal rather than exceptional. The skill barrier keeps dropping, the tooling keeps forking, and the economics favour volume. Organisations have significant cyber debt that is easily exploitable by frameworks like Artex, so we should expect persistent disruption of digital services – travel, banking, commerce, etcetera – in the coming months as these organisations scramble to quickly improve their cyber resilience.”

Antani warned that any defenders still inclined to treat exposure management as a quarterly scanning or annual penetration testing exercise will find themselves swiftly overwhelmed.

“Machine-speed offence is already here; machine-speed defence has to match it,” he said.



Source link