On November 16, the FBI issued its latest cybersecurity advisory, warning that hackers are using Artex, an open-source AI penetration-testing tool, to launch autonomous attacks — in some cases going from targeting a system to actively compromising it in under three hours. The bureau says several industries have already been hit in recent months, including finance, government, tech, defense, telecoms and law firms, and that attack speeds have now outpaced the rhythm of human defense.
How Artex breaks into networks
Artex was built to help companies probe their own network defenses. It can connect to frontier large language models from Anthropic and OpenAI, as well as to open-source and non-U.S. models. Hackers use it to scan a target network quickly, find simple flaws such as weak passwords or exposed credentials to gain a foothold, then spread sideways through databases, mailboxes and other sensitive information. The FBI concedes that such intrusions are 'noisy' — they generate heavy traffic and are easy to trip alarms — but the speed advantage often means defenders arrive after the attack is already over.
The damage so far
Artex first appeared publicly in July, and has since been used in several real attacks. A number of South Korean financial institutions were hit, with the personal data of at least 68,000 people stolen; weak defenses, including missing multi-factor authentication, were largely to blame. Customers of ConnectWise, the American IT management company, were also breached: the Microsoft 365 cloud environments of at least 2,250 customers were accessed, and Microsoft says the attackers were likely Chinese. Cybersecurity firm Barracuda was hit too: the attackers got in with a credential belonging to a third-party service that should have been revoked. Unable to establish the full scope from its logs, the company chose to 'assume the worst' and rotate all of its customer credentials.
After the attacks came to light, Artex's developer made the software closed-source — but the FBI says the earlier open-source version keeps circulating and remains widely accessible. Last week, researchers found hundreds of Artex systems exposed on the internet without a password, including the one used to target South Korea. Anthropic recently cut its vulnerability disclosure window from 90 days to 30 days on the grounds that AI finds flaws too fast: the window defenders have is shrinking. For companies, the practical steps are concrete — turn on multi-factor authentication, revoke credentials for services you no longer use, and do not just watch alerts and wait for an attack to happen. With a three-hour window, by the time you notice, it is already too late.