FBI: ARTEX AI framework used to breach networks in under three hours — crypto firms targeted The FBI has warned that attackers are using an AI framework called ARTEX to break into networks with alarming speed — sometimes gaining privileged cloud access in under three hours. In a Nov. 13 advisory, the bureau said ARTEX-driven intrusions have been seen across financial services, government, defense, healthcare and technology suppliers since at least August. One incident cited involved the theft of personal information for roughly 67,000 people in South Korea. What ARTEX does - Built for authorized penetration testing, ARTEX lets operators give natural‑language instructions to AI agents. Those agents then execute multi-step tasks that previously would have required manual attacker effort, automating reconnaissance, credential hunting and lateral movement. What researchers and vendors found - Microsoft (Oct. 21): Attributed a campaign to Storm‑0940, a group Microsoft says is aligned with Chinese state actors. The company said the group used ARTEX against hundreds of organizations, and some intrusions reached privileged cloud access within the FBI’s three‑hour window. Following a ConnectWise compromise, attackers used stolen application credentials to attempt sign-ins to more than 2,250 Microsoft 365 tenants — a figure that describes attempted access rather than confirmed, successful breaches. In a separate password‑spraying run across 327 accounts in eight organizations, Microsoft recorded more than 151,000 failed logins and observed no successful authentications. - Google (Oct. 30): Reported ARTEX activity in the United States and Europe targeting semiconductor, telecommunications and enterprise‑software companies. Researchers saw both financially motivated operators and suspected espionage groups. Although ARTEX can move quickly, Google also observed repeated errors and heavy logging; a single intrusion generated about 240 log events, and separate agents sometimes interfered with one another. - Barracuda (Nov. 11): Traced an intrusion to a cloud access key it had supplied to an AI vendor and failed to retire after the vendor’s work. After the vendor’s development environment was compromised in October, an attacker used the key to enter Barracuda’s cloud on Nov. 1. Over roughly 44 hours the intruder reached about 10 cloud accounts by following credentials from system to system. Barracuda said the sequence and built‑in credential hunting routines matched ARTEX patterns but stressed the tool alone did not identify the attacker. The intruder searched compromised servers for Ethereum keystores, extended private keys and seed‑phrase files; Barracuda reported none were present and acknowledged the lapse in revoking vendor access. - Amazon (Nov. 13): Observed crypto‑related campaigns that tried to bypass identity verification at two exchanges using deepfake video and altered identity documents. Agent records sent to Amazon’s decoy systems between Oct. 3 and Nov. 5 implicated ARTEX with moderate confidence. One verification vendor rejected every deepfake attempt recorded, but attackers ran simultaneous agent sessions for video generation, verification, and other account tasks. Amazon also saw campaigns that harvested credentials for AI services and reused stolen keys to pay for further AI processing. The company noted attackers exploited familiar weaknesses: exposed credentials, unpatched systems and unauthenticated interfaces. - Other reports: GitBook notified customers on Nov. 3 to rotate integration secrets after an AI‑agent attack accessed infrastructure credentials (API tokens, Git Sync secrets, and data submitted to GitBook Assistant). In South Korea, Everspin reported ARTEX‑related attempts against financial services while announcing an AI identity partnership with SBI Holdings and KB Kookmin Card. Why crypto is a special risk - Attackers searched for wallet material (keystores, private keys, seed phrases) and tried to defeat KYC systems with deepfakes — direct threats to exchanges, custody providers and any organization that stores or processes crypto keys or handles on‑chain transactions. - Attackers also recycled stolen AI‑service credentials to fuel further automated attacks, creating a self‑sustaining attack infrastructure. Practical steps for defenders (short checklist) - Revoke and rotate vendor keys and integration secrets immediately after engagements. - Enforce least privilege, strong identity controls and multi‑factor authentication for cloud and admin accounts. - Monitor for high‑volume/hyperautomated behavior and unusual API or agent activity in logs. - Protect and isolate wallet material; treat keystore files and seed phrases as high‑risk artifacts. - Harden identity verification systems against deepfakes (liveness checks, multi‑factor KYC) and keep software patched. - Audit third‑party vendor security practices and limit long‑term access where possible. Bottom line ARTEX shows how AI automation can accelerate traditional attack chains and scale reconnaissance and credential harvesting — and that speed matters. Crypto firms and financial institutions are squarely in the crosshairs because successful breaches can directly translate to stolen funds or compromised customer identity. The FBI and multiple vendors urge immediate attention to vendor access, credential hygiene and hardened KYC to blunt these AI‑driven campaigns. Read more AI-generated news on: https://app.chaingpt.org/news