Key facts
- Barracuda confirmed it was hacked in an intrusion that unfolded between November 1 and November 2, 2026.
- The FBI says ARTEX lets minimally skilled threat actors run complex operations against many targets at once, speeding up attacks and expanding their scale.
- Trackers listed Barracuda as hacked on 16 November 2026 at 15:18 UTC.
- The attacker reused discovered secrets to enter additional systems and take more credentials and secrets.
Barracuda confirmed an attacker hacked its cloud environment in an intrusion that ran from Nov. 1 to Nov. 2, 2026, after the attacker used an exposed valid cloud access key to map the company’s reach, and breach trackers listed the incident on Nov. 16, 2026 at 15:18 UTC.
After gaining access, the intruder reused discovered secrets to enter additional systems and take more credentials and secrets, reaching about 10 cloud accounts over roughly 44 hours, Barracuda said. The company discovered the intrusion on Nov. 4, 2026, forty-eight hours after the attacker’s last activity, alerted by its AI provider and by an engineer who noticed unusual activity on their own account. The attacker deployed ARTEX to escalate access across Barracuda’s internal network and other cloud environments and moved through 10 cloud environments, including some running Barracuda’s extended detection and response platform.
Barracuda found no evidence the attacker accessed extended detection and response customer integration credentials that could let the attacker pivot into customer networks. With shell access on some servers, the intruders searched for extended private keys, Ethereum keystore files, seed phrases, and other wallet-related information. They targeted Kubernetes service accounts, server metadata services, secret stores, and configuration files, and planted a backdoor inside a Kubernetes cluster.
Barracuda has not assigned responsibility for the breach because many groups use ARTEX, though the company also believes a financially motivated group carried out the intrusion. Its probe found that the incident’s names, indicators, system access sequence, and credential- and wallet-hunting routines match ARTEX. Barracuda said the attack resembled the October intrusion at cloud application hosting platform Railway and fit a wave of intrusions at dozens of AI developer platforms.
ARTEX is an agentic AI penetration testing framework that a developer known as Autumn-27 published on GitHub in July 2026. GitHub removed its code in October 2026, but the framework keeps spreading through other channels. It works through a planner that builds objectives and tasks, agents that carry out the tasks, and a platform that records progress, findings, and assets.
SecurityWeek reports that ARTEX can use large language models, reportedly beginning with DeepSeek v4.1-flash, and that GLM-5.3 and Grok 4.6 have appeared in certain instances. The outlet also says users mainly rely on Chinese models, which are more permissive than U.S.-made ones.
Microsoft says ARTEX agents default to Nmap, curl, wget, Python HTTP clients, Norma SDK WebFetch, and headless Chromium via Playwright. They can scan and exploit web application flaws, persist with web shells, steal credentials, and abuse compromised cloud identities to keep access and export data, and they can run brute-force and password-spray attacks, list and abuse APIs, and dump databases. After compromise, ARTEX-linked actors misuse credentials and keys, raise privileges, maintain persistence, move sideways, and steal credentials and data, SecurityWeek reports.
CrowdStrike attributes Vapor Panda to a China-based intrusion cluster first identified in February 2021 and formerly known as VexatiousTemper. Starting Oct. 17, 2026, the espionage group used ARTEX against more than 200 organizations in 14 countries, including foreign ministries, a European presidency, and U.S. pharmaceutical, semiconductor, defense, and nuclear organizations.
Microsoft says Storm-0940 is a China-state-aligned actor that used ARTEX to speed up sign-in attempts against more than 2,250 downstream Microsoft 365 tenants using compromised ConnectWise application credentials. In an October supply-chain attack, ConnectWise Cloud Backup compromised Microsoft 365 customer environments through application identities formerly tied to SkyKick Cloud Backup, and Huntress identified at least 162 compromised tenants.
Storm-3358, a Cambodia-based intrusion cluster, began targeting financial organizations in October 2026 using ARTEX, mainly in Southeast Asia plus a small set of large financial organizations worldwide, Microsoft says. It used WISP, vshell, and AdaptixC2 remote access tools, which ARTEX agents downloaded with netcat and curl. Cisco’s Talos says financially motivated actors including Storm-3358 have adopted the framework.
The FBI says ARTEX lets minimally skilled actors run complex operations against many targets at once, speeding up attacks and expanding their scale. The FBI says threat actors have used ARTEX since at least August 2026 against sectors including finance, government, and healthcare, and that some intrusions reach privileged access in under three hours from initial targeting.
A Chinese-speaking actor paired ARTEX with Claude Code agents to attack South Korean banks including Shinhan Bank, KB Kookmin Bank, and Hana Bank, exposing personal information of about 67,000 people. South Korean police arrested seven suspects tied to attacks on more than 90 organizations, with ARTEX reportedly used in most intrusions, and authorities say the gang stole about 8.2 million data records from hospitals, educational institutions, and businesses.
Researchers identified more than 1,600 ARTEX consoles exposed to the public in early November 2026, while Censys saw roughly 1,500 hosts running exposed consoles in the same period. Amazon says one major campaign stole credentials at scale to reach AI models and another hit cryptocurrency exchanges, with attackers using well-known methods but moving much faster because of AI. SentinelOne puts the typical cost of a campaign at $700 to $1,500, spanning e-crime to cyber-espionage, and AWS sees Chinese-nexus groups abusing ARTEX for mass exploitation, including defeating know-your-customer (KYC) checks at cryptocurrency services. SecurityWeek says real-world use of the framework has risen in recent weeks.
Cisco Talos reported on Nov. 13, 2026, that malicious use of AI penetration testing frameworks including ARTEX is growing, tied to credential theft and malware deployment. It advises linking authentication, endpoint, and cloud API logs to detect fast requests, unknown applications, abnormal token usage, and privilege changes, and recommends patching exposed systems, removing default passwords, enforcing multifactor authentication, applying rate limits, scanning public repositories for secrets, and rotating exposed credentials. The FBI urges organizations to secure internet-facing applications, restrict cloud administration access, review application permissions and authentication logs, apply rate limits on sensitive APIs, and watch for sudden waves of automated requests, odd cloud behavior, unexpected credential or privilege changes, and large data transfers.
Microsoft provided indicators of compromise including the IP addresses 199.19.95.66, 199.19.95.67, and 199.19.95.68 and 91.92.240.238, 85.11.168.45, 178.15.54.129, 31.77.229.35, and 85.11.167.52, and it said defenders should use AI across large environments to investigate and respond as agentic framework abilities develop. MalwareMustDie released an open-source ARTEX detection signature pack covering network requests, identity misuse, automated attack patterns, cloud activity, and malicious files.
Amazon says fixing the basics — default or hardcoded credentials and unpatched software — is the most effective defense against ARTEX. Daniel Shechter of Miggo Security says defenders should shift from spotting exposure to proving protection at machine speed. Risky.Biz notes that a breach weighs more heavily when the victim is a security provider, because such firms hold privileged access to customer environments and build reputations on not being compromised. Barracuda notified some customers and advised them to reissue those credentials as a precaution.

Join the discussion