The agency says criminals and state-linked groups have exploited the open-source tool since August 2026, targeting multiple sectors and reaching privileged access within hours
The FBI has warned that attackers are increasingly using ARTEX, an open-source, AI-driven penetration-testing framework, to automate intrusions at “machine speed.”
According to Bleeping Computer, the agency had issued a FLASH alert on 13 November 2026, saying malicious actors have used ARTEX since at least August 2026 against organizations in financial services, government, defense, energy, healthcare and other sectors.)
Some intrusions had progressed from initial targeting to privileged access in under three hours. ARTEX does not include its own AI model; it connects to external providers including Anthropic’s Claude, OpenAI’s GPT and DeepSeek.
The framework has also been linked to a recent breach of cloud-security firm Barracuda. An attacker had entered parts of the firm’s cloud environment between 1 November 2026 and 2 November 2026, reaching about 10 cloud accounts in roughly 44 hours. The firm discovered the intrusion on 4 November 2026.
Breach incident details
The attacker had used a valid cloud access key that the firm had issued to a third-party AI vendor for a completed project. Barracuda said it did not retire the key when the work ended, acknowledging that the lapse was its responsibility. It believes the key had been stolen in October 2026 from a configuration file in the vendor’s GitHub repository. Then:
- Using credentials found in one system, the intruder had moved into the firm’s data analytics environment, source-code repositories and Kubernetes clusters.
- According to the firm, the attacker’s probe names, sequence of activity and automated routines for finding credentials and cryptocurrency wallets matched ARTEX.
- The attacker then searched for Ethereum keystore files and seed phrases, and installed a backdoor in a Kubernetes cluster. It had also accessed a system containing integration credentials for the firm’s XDR managed threat-detection service.
- The firm has yet to find evidence that those credentials were stolen, but has asked potentially-affected customers to revoke and recreate them.
Barracuda has not attributed the intrusion to a particular actor.
Growing abuse
Although ARTEX’s original developer had removed its GitHub repository on 8 October 2026, the FBI has said the tool continues to circulate through community forks, localized variants, archived copies and a commercial version. In early November 2026 researchers at Censys had identified more than 1,600 publicly exposed ARTEX consoles.
The American Hospital Association has urged hospitals to inventory every internet-facing application. John Riggi, its national advisor for cybersecurity and risk, said ARTEX enables both skilled and unskilled hackers to operate at “machine speed.”
On 16 November 2026, Google said its threat analysts had observed several threat actors using ARTEX and similar frameworks in live intrusions since late August 2026. The group also named CAIRN, CyberStrikeAI, HexStrike AI and PentAGI as other open-source, multi-agent frameworks being abused, according to this report.)
