The FBI warned on Nov. 13 that autonomous, AI-driven hacking tool Artex is enabling intrusions that move from initial targeting to privileged network access in under three hours, outpacing defenders’ ability to respond, according to Science Et Vie.
The alert, issued as a FLASH bulletin, documents global campaigns since summer hitting finance, defense, technology, energy, semiconductors, telecommunications, healthcare, government and legal firms in the United States and Asia. Since summer, “divers cyberacteurs malveillants ont utilisé ARTEX pour mener des campagnes visant des organisations du monde entier,” the FBI said, according to 01Net.
Artex chains commercial or open large language models — including Anthropic’s Claude, OpenAI’s GPT, and Chinese models such as DeepSeek, Kimi and GLM — to automated scanning and exploitation modules, allowing operators with minimal technical skill to run complex, simultaneous operations. According to 01Net, the FBI said Artex “reduces considerably the time needed to chain attack techniques, some intrusions going from initial targeting to privileged access in less than three hours,” and that “the speed of the attack leaves little time for a coordinated response by defenders before the objective is reached.”
The bureau added that Chinese models, particularly the DeepSeek Flash family, are favored for their “relatively low cost” and the absence of “observable guardrails when the request is presented as a penetration test.” To delegate the work, Artex first translates the user’s objectives into “commandes exécutées par des agents spécialisés,” after which only “interaction humaine minimale” is required.
This week in FBI Cyber: From disrupting malicious cyber actors to defending against threats, FBI Cyber is taking action to protect the American people. Here’s a look at some of our work this week: ➡️Following North Korea’s theft of over $6 billion in virtual assets from… pic.twitter.com/GNMeteYNCQ
Developer “Autumn-27” published Artex on GitHub in July as a legitimate penetration-testing platform, then closed the repository on Oct. 8 after early abuse. Despite that takedown, the FBI bulletin says the code remains “widely accessible.” Science et Vie reports that researchers found hundreds of exposed, passwordless Artex consoles online as of early November, with more than 1,600 still accessible worldwide. Microsoft discovered that many pirate servers identified by the FBI remain active.
Two confirmed campaigns illustrate the impact. Suspected Chinese actors used Artex to breach ConnectWise, a U.S. IT management platform, gaining access to the Microsoft 365 environments of at least 2,250 clients. In South Korea, the same tool compromised multiple financial institutions that lacked multifactor authentication, stealing personal data of at least 68,000 individuals. L’informatique notes that Google’s Threat Intelligence Group (GTIG), in a report released Nov. 16 with Mandiant, separately documented the shift from manual prompting to fully autonomous agent chains — citing not only Artex but also platforms CAIRN, PentAGI and CyberStrikeAI — capable of chaining reconnaissance, exploitation and lateral movement without human intervention.
GTIG also observed automated reverse-engineering of patches to weaponize known vulnerabilities on exposed servers, and “LLMJacking” campaigns that steal cloud API keys to resell access to advanced models via covert proxies.
The FBI confirmed the tool was used to hack South Korean financial institutions, resulting in the theft of personal data concerning at least 68,000 people, as well as the publisher ConnectWise with illicit access to Microsoft 365 environments of at least 2,250 clients. Recently targeted sectors include financial services, administrations, IT publishers, defense, telecoms and law firms, according to the federal alert.

The FBI bulletin recommends reducing internet-facing attack surfaces, enforcing multifactor authentication and revoking exposed cloud credentials. Google GTIG adds that rapid patching is essential, as the speed of Artex-driven intrusions — noisy though they may be — leaves on-call teams with almost no maneuvering room before data exfiltration begins. On the technical level, intrusions conducted with Artex remain noisy and generate an unusual volume of network requests likely to trigger detection probes.
But the rapid chaining between the first scan, exploitation of stolen passwords or the absence of multifactor authentication and the pillaging of mailboxes leaves very little room for maneuver for on-call teams.
This race against the clock also affects the internal software supply chain, since Google mentions the manipulation of code assistant configurations to deposit backdoors, a risk already illustrated by a sandbox escape. To contain these offensive agents, rapid application of patches, generalization of multifactor authentication and revocation of exposed cloud keys become indispensable.
The #FBI issued a FLASH advisory warning that cybercriminals and state-aligned hackers are exploiting ARTEX, a super intelligence-powered penetration testing framework, to automate attacks against businesses and organizations worldwide. ARTEX can accelerate intrusions from… pic.twitter.com/TwsbGYJfH4
Related reading
- Summit Explores Post-Von Neumann Computing From Superconductors to Biology
- Some Players Find Vice City Least Interesting Part of GTA VI
- Ex-NANS President Warns Student Loans Cannot Replace Public Education Funding (archynewsy.com)
- November Beaver Moon Is First of Three Consecutive Supermoons (time.news)