The FBI has issued a flash cybersecurity advisory warning that hackers in the United States and globally are using an open-source AI-powered hacking tool called Artex to launch autonomous cyberattacks — in some cases moving from initial targeting of a system to active compromise in under three hours. The advisory was reported by TechCrunch on November 16, 2026.

The campaigns cited in the advisory include a widespread hack of customers of American IT management company ConnectWise and attacks across South Korea's financial industry. It is the clearest signal yet from US law enforcement that autonomous offensive AI tools have moved from research demos to working break-in kits. We have been tracking this escalation in our breaking AI news coverage since the summer, and the FBI's advisory now puts official weight behind what security researchers have been saying for months.

What the FBI Says About Artex

According to the advisory, Artex is one of several open-source AI-powered penetration testing platforms — a category of software originally built to probe a network's defenses. The difference is who is pointing it at whom. The FBI says hackers have connected Artex to hosted frontier large language models from companies including Anthropic and OpenAI, as well as to open-source models and models based outside the United States.

The agencies and companies targeted in recent months span a wide swath of the economy: financial services, government, technology, defense, phone and internet companies, and law firms. The FBI notes that Artex has been used in multiple cyberattacks since the tool first appeared publicly in July.

From Target List to Foothold in Three Hours

The headline statistic in the advisory is speed. In some cases, the FBI says, Artex-linked attacks moved from targeting a system to actively compromising it "in under three hours." That is a timeline most human-staffed security operations cannot match — aSOC analysts still have to notice an alert, triage it, and respond, while an autonomous agent never sleeps and never queues.

The FBI also offers a candid caveat: Artex intrusions can be "noisy," generating reams of traffic and data that may trip a targeted company's defensive alerts. But volume is not the same as velocity. The tool typically works by rapidly scanning a victim's network for signs of weakness, then exploiting simple security defects — weak protections, exposed credentials, or stolen passwords — to gain a foothold. From there, the FBI says, it spreads across the network in search of databases, mailboxes, and other sensitive information.

ConnectWise and Barracuda: The Case Studies

The most consequential confirmed incident involves ConnectWise, whose software is used by thousands of IT service providers to manage client systems. The FBI confirmed earlier findings from Microsoft that hackers used Artex to break into the company and gain access to the Microsoft 365 cloud environments of at least 2,250 of its customers. Microsoft assesses that the hackers are likely Chinese and used the tool to steal data from affected customers.

Security company Barracuda also disclosed a breach connected to Artex, attributed to financially motivated criminals. In that case, the AI tool found an exposed credential belonging to a third-party service that Barracuda had not revoked after ending work with the vendor. That single leftover credential gave hackers access to stores of credentials Barracuda used to collect data from its customers' systems.

Barracuda's chief information security officer, Arve Kjoelen, wrote that the company could not rule out data access because its logs were not detailed enough, and chose to "assume the worst" and rotate all of its customer credentials. "Defenders who rely on spotting malicious activity and then responding to it will often arrive after the attacker has finished," he wrote. A company spokesperson told TechCrunch that Barracuda "directly notified and advised a minority of XDR customers to revoke and recreate integration credentials," but did not specify how many.

South Korea Was the Warning Shot

For readers following this story, none of the FBI's warnings will come as a complete surprise. We reported in October on South Korea's disclosure that AI was used in bank hacks exposing data at seven financial firms, where attackers stole the personal data of at least 68,000 people. Investigators blamed weak cybersecurity defenses and missing protections such as multi-factor authentication as much as the tooling itself.

The pattern repeated in a Japanese-led probe we covered just last week, which traced 1,000 servers powering AI-driven cyberattacks worldwide. The FBI advisory now confirms the same tool at the center of both storylines was also hitting US targets.

Closed-Source Now, But the Genie Is Out

There is one twist worth noting. After Artex was used in real attacks, its developer made the software closed-source to prevent further abuse. The FBI's assessment, however, is that the previous open-source version continues to circulate and remains "widely accessible" for use in cyberattacks.

The exposure problem extends to the tool's own infrastructure. Last week, researchers found hundreds of publicly exposed Artex systems sitting on the open internet without password protection — likely through misconfiguration — including the very Artex system used in the South Korea attacks. Offensive AI tooling, in other words, is now itself an attack surface.

What Defenders Should Take Away

The lesson of the advisory is less about Artex specifically than about what it represents: the marginal cost of executing a competent network intrusion is collapsing. The incidents the FBI describes were not enabled by exotic zero-days. They succeeded on rapid scanning, unrevoked vendor credentials, missing multi-factor authentication, and exposed passwords — the same fundamentals human-driven attackers have exploited for decades, now executed at machine speed.

For defenders, the priorities the cases suggest are familiar but newly urgent: enforce multi-factor authentication everywhere, audit and revoke third-party integration credentials when relationships end, and assume that detection will lag compromise. As Barracuda's CISO put it, arriving after the attacker has finished is not a defense strategy — it is a breach report.

Stay Ahead of AI

The AI landscape moves fast, and the stories that matter most often break between the headlines. Follow our AI news coverage for reporting you can verify.

Read more AI news →