ARTEX: FBI Identifies Breaches, AWS Details Agents

From South Korean banking data to Microsoft environments of indirect clients, the FBI reports several compromises associated with ARTEX. AWS, on its part, describes agents employed to steal access to AI models or attempt to bypass identity controls, with varying results depending on the operations.
Censys identifies 1,465 accessible ARTEX consoles
As of November 4, Censys had identified 1,465 hosts leaving their ARTEX console accessible. The FBI estimates that the actual deployment is likely much broader, with installations protected by firewalls or hosted on private networks escaping this count. The removal of the GitHub repository by the initial developer in October, following the disclosure of the compromises, did not eliminate the tool: derived variations, localized versions, and a commercial offering persist.
The FBI specifically cites South Korean financial institutions attacked in August, with personal data stolen concerning approximately 67,000 individuals. Another reported case involved an ARTEX operation conducted in September against clients of ConnectWise Cloud Backup. This supply chain attack reportedly led to the compromise of over 2,250 downstream Microsoft tenants, according to the federal agency.
The FBI recommends restricting sensitive access
In its separate alert issued on Friday, the FBI indicates that some intrusions achieved privileged access in less than three hours after the initial targeting. ARTEX accelerates, according to the agency, the sequence of offensive techniques. For Amazon Threat Intelligence, the methods employed are not new; the contribution of the agents lies in the speed and persistence of the operations.
CJ Moses attributes the observed successes to basic shortcomings: default or hardcoded credentials, keys left in code or public elements. He also mentions interfaces and inference endpoints lacking authentication, as well as unpatched software. The official considers mastering these fundamentals as the most effective protection.
The FBI urges all organizations to implement the recommendations from its Flash alert, particularly those using applications exposed on the Internet, self-service portals, cloud environments, or systems belonging to partners and suppliers. It advises limiting access to sensitive services to networks deemed trustworthy, private links, or specific administrative pathways. Enterprise applications should also be subject to policies and permissions based on a zero-trust model.
Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisissez votre rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
AWS describes key theft fueling other agents
The FBI and Amazon Threat Intelligence report the offensive use of ARTEX, an open-source framework intended for penetration testing, against banks, government agencies, and organizations in the energy or healthcare sectors, among other critical areas worldwide. Amazon's MadPot sensors detected two campaigns between October 3 and November 5. One was heavily seeking credentials to consume AI models without payment.
The agents stole API keys, which were then injected into the ARTEX model gateway to make other agents work in search of additional access. CJ Moses, AWS's Chief Information Security Officer, detailed their operations on Friday: validating identities with identity and token services from providers, inventorying storage, and reviewing permissions. According to him, the targets included managed model services, secrets, parameter stores, and application identities; cloud credentials were also used to access the models.
Among the targeted LLM gateways and proxy platforms were One API, New API, Veloera, Sub2API, and LiteLLM. The campaign also targeted AI application creation tools: Dify, FastGPT, Langflow, RAGFlow, Open WebUI, LobeChat, and Ollama.
Server control rejects synthetic faces
The other campaign observed by Amazon aimed to steal cryptocurrencies from two exchange platforms. The attackers used ARTEX to create deepfake videos in real-time against their KYC controls, including live selfies, and falsified identity photos intended for document verification. According to CJ Moses, several sessions operated simultaneously, with agents dividing tasks such as finding withdrawal methods, cleaning account balances, producing videos, and bypassing checks.
The agents had reverse-engineered the detection protocol of a vendor, down to facial detection events, face tracking, and coordinates guiding the movements requested from the user. They completed the gesture sequence with a face placed within 7 pixels of the target. Despite this precision, the server rejected the deepfake on every attempt of this sequence.
In a separate session, a video chain based on an open-source facial animation model operated on a single consumer GPU. It achieved 11.5 frames per second, a rate deemed by Moses to be close to the expectations of a live video check. For documents, the agents fabricated camera metadata consistent with the actual dimensions of the JPEG and preserved the original compression tables, which are typically replaced when a retouched image is saved again.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.