Federal investigators report that cybercriminals are deploying the ARTEX artificial intelligence framework to breach corporate networks in under three hours. The agency notes that threat actors are leveraging this tool to secure high-level system control, primarily targeting financial firms and technology vendors.
A Nov. 13 federal bulletin details how attackers have deployed ARTEX against banking, government, defense, and healthcare sectors since at least August. Officials highlighted one major breach where personal data belonging to roughly 67,000 South Korean citizens was stolen.
Originally designed for legitimate security audits, the platform allows users to direct artificial intelligence agents through plain conversational prompts. These autonomous tools then execute complex sequences that would normally demand tedious manual effort from an intruder.
ARTEX Attacks Abuse Trusted Access to Infiltrate Customer Networks#
An Oct. 21 probe by Microsoft linked a widespread campaign to Storm-0940, a China-affiliated threat actor that deployed ARTEX against hundreds of enterprises. In several breaches, administrative cloud control was obtained within the three-hour timeframe the FBI previously flagged.
In the wake of the ConnectWise breach, threat actors leveraged stolen application credentials to attempt logins across over 2,250 Microsoft 365 environments. Subsequent malicious activity was detected by Microsoft within several client networks, encompassing system reconnaissance alongside unauthorized alterations to app credentials, user identities, and administrative permissions.
That tenant metric reflects intrusion attempts rather than 2,250 verified compromises. During a separate password-spraying campaign targeting 327 accounts across eight enterprises, over 151,000 unsuccessful login tries were logged by Microsoft with zero confirmed breaches.
An Oct. 30 report from Google detailed ARTEX intrusions targeting enterprises across the United States and Europe, spanning semiconductor, telecom, and business software firms. Analysts noted that the framework was deployed by both profit-driven cybercriminals and suspected state-backed espionage operators.
Despite the rapid pace of these operations, Google uncovered frequent glitches and extensive audit trails. During a single breach, 240 log events were produced by the automated tools, showing clear evidence that competing agents disrupted each other’s tasks.
Barracuda Hacker Searched for Cryptocurrency Wallet Data During Attack#
A Nov. 11 report from Barracuda detailed a security breach stemming from a cloud access key earlier provided to an AI vendor. Officials noted that the credentials had not been revoked after the contractor wrapped up its assignment.
Following an October breach of the contractor’s development system, the intruder reportedly acquired that credential and leveraged it to infiltrate Barracuda’s cloud infrastructure on Nov. 1.
Stay in the loop
Get crypto news before the market moves
Join thousands of investors who read our daily briefing.
No spam. Unsubscribe anytime.
Across roughly 44 hours, the threat actor accessed nearly 10 cloud accounts by tracing credentials from one environment to the next. Barracuda noted that the operational sequence, digital footprints, and automated credential-harvesting routines aligned with ARTEX, while emphasizing that the software alone could not be used to pinpoint the culprit.
Inside breached servers, the intruder hunted for Ethereum keystore files, extended private keys, and seed phrase documents. Barracuda confirmed that no cryptocurrency wallet data was stored on those machines.
The firm owned up to its oversight in failing to terminate the contractor’s permissions, stating bluntly that the mistake was on them.
Amazon Discovers Cyberattacks Targeting Crypto Exchange Identity Checks#
A Nov. 13 report from Amazon detailed a separate cryptocurrency-focused operation where attackers tried to circumvent identity checks at two platforms, utilizing synthetic deepfake footage alongside forged identification papers.
The intelligence emerged from agent logs transmitted to Amazon’s honeypot systems between Oct. 3 and Nov. 5. Researchers attributed the campaign to ARTEX with moderate confidence, noting that intrusions against every entity named in those files could not be independently verified.
For one screening provider, every documented deepfake submission was turned down by the server, even when the automated tool executed the mandatory facial-motion prompts. Independent agent instances operated concurrently to handle video creation, identity checks, and additional account procedures.
A separate operation harvested logins for artificial intelligence platforms and routed those stolen keys back into the hackers’ systems to fund ongoing AI workloads. Amazon noted that these tactics depended on common security flaws, such as leaked credentials, outdated software, and unprotected interfaces.
Credential Theft Prompts Customer Account Resets Across Other Platforms#
As previously covered on Nov. 3, GitBook urged clients to rotate their integration secrets after an AI-agent intrusion compromised underlying infrastructure credentials. Potentially exposed records included API tokens, Git Sync keys, and user prompts that were submitted to GitBook Assistant.
In South Korea, Everspin reported spotting ARTEX-linked intrusion attempts within financial sector networks as it unveiled an October AI identity venture with SBI Holdings and KB Kookmin Card. Before automated payments are rolled out commercially, the partners intend to evaluate user consent, spending caps, and delegated permissions.



