
Hackers found a way to break into company networks in under three hours using an open-source artificial intelligence tool, the Federal Bureau of Investigation (FBI) says.
Quick Take
- The FBI warns that a tool called ARTEX has helped hackers attack organizations across finance, government, defense, energy, and healthcare since at least August 2026.
- The bureau says some break-ins went from first contact to deep system access in under three hours.
- Security firm CrowdStrike tracked one ARTEX operator hitting targets in 14 countries over 11 days, confirming real damage at nine organizations.
- The FBI’s Dallas office says related attacks stole personal data on roughly 67,000 people from South Korean financial institutions.
What The FBI Says Happened
The FBI issued a FLASH advisory warning that ARTEX, an open-source artificial intelligence platform, has become a weapon for cybercriminals and state-linked hackers. The bureau says the tool has targeted financial services, government agencies, technology firms, defense contractors, semiconductor makers, energy companies, hospitals, manufacturers, telecom providers, and law firms since August 2026. That is not a narrow problem. It spans nearly every sector that keeps the American economy running.
The most alarming detail is speed. The FBI says ARTEX “significantly reduces the time required to chain attack techniques,” with some intrusions moving from initial targeting to privileged access in under three hours. For context, traditional hacking campaigns often take days or weeks of manual probing before an attacker gains that kind of foothold. If the FBI’s timeline holds up, defenders have far less warning than they used to.
ARTEX itself is not a self-contained artificial intelligence brain. Reporting describes it as an agentic framework that plugs into outside large language models, including Claude and ChatGPT, to automate steps from reconnaissance through data theft. In plain terms, ARTEX is the orchestra conductor, and it borrows the instruments from elsewhere. That distinction matters because it means the danger is not one rogue program but a method that can be rebuilt with whatever model is available.
Outside Researchers Back Up The Warning
The FBI is not alone in flagging this activity. CrowdStrike’s threat-hunting unit reported that starting October 17, 2026, a single ARTEX-driven controller directed attacks against organizations in at least 14 countries over just 11 days. CrowdStrike confirmed that attackers actually broke into and operated inside nine of those organizations, deploying malware known as BLOODALCHEMY and then stealing passwords and mapping networks. That is hard evidence of real damage, not just attempted break-ins.
The FBI’s Dallas field office tied ARTEX-linked attacks to a separate, concrete harm. Hackers using the tool stole personal information belonging to roughly 67,000 people during attacks on South Korean financial institutions in late September and early October 2026. Numbers like that turn an abstract warning about artificial intelligence into a story about real people whose data is now in criminal hands.
Fitting Into A Longer Trend, Not A Sudden Leap
None of this emerged out of nowhere. Security researchers have spent years building and studying automation frameworks that script out attack chains, from scanning networks to stealing credentials, cutting the manual labor hackers once needed. ARTEX fits that same progression, just with a large language model doing more of the thinking. The real shift is not that machines attack alone, but that the gap between a skilled human hacker and an automated one is shrinking fast.
The FBI’s advisory should be read as a serious operational warning rather than a one-time news cycle. It reportedly was not even a brand-new announcement when it drew fresh attention, suggesting the bureau has been tracking ARTEX activity for some time. That is a point worth sitting with: law enforcement saw this pattern building before most of the public ever heard the name ARTEX.
Why This Should Worry Every American Business
American companies, hospitals, and utilities now face a threat that moves at a pace human defenders were never built to match. A three-hour window from first contact to privileged access leaves almost no room for a security team to notice, investigate, and respond before real damage is done. That reality should push every organization running critical systems to treat artificial intelligence-driven threats as a baseline risk, not a future hypothetical.
🚨💀 FBI: AI HACKING TOOL BREACHES NETWORKS IN UNDER 3 HOURS
ARTEX is an open source AI agent framework that autonomously plans and executes attacks at machine speed.
Already used against Korean banks (67k records stolen) and ConnectWise (2,250+ MS365 tenants).
Hundreds of… pic.twitter.com/p6HpAEz4DD
— OVERCLASSIFIED (@overclassifiedx) November 18, 2026
Open-source tools built for legitimate security testing can be turned against the very systems they were meant to protect, and ARTEX is the latest proof. The FBI’s warning gives security teams a name and a pattern to watch for. Whether that warning triggers real changes in how fast American companies patch systems and train staff will matter far more than the headlines about the tool itself.
Sources:
abcnews.com, dallasnews.com, washingtonexaminer.com, crowdstrike.com, leadershipsummit.aha.org, therecord.media














