The US Federal Bureau of Investigation (FBI) has warned that hackers are using autonomous artificial-intelligence tools to break into organisations in the United States and other countries, and that in some cases an attack moved from picking a target to an active compromise "in under three hours". The warning came in a flash cybersecurity advisory reported by TechCrunch on 16 November 2026.
The advisory centres on Artex, an AI-powered penetration-testing platform. Such tools are built to probe a network's defences, and Artex can be connected to large language models from companies such as Anthropic and OpenAI as well as open-source and non-US models, TechCrunch reported.
Who is being targeted
According to the report, the FBI listed financial services, government, technology, defence, phone and internet companies and law firms among the sectors hit.
TechCrunch described three incidents linked to the tool:
- Several South Korean financial institutions were breached, largely because of weak defences and missing multi-factor authentication. Personal data of at least 68,000 people was stolen.
- US IT firm ConnectWise was breached and the attackers reached the Microsoft 365 environments of at least 2,250 of its customers, according to earlier Microsoft findings cited in the report. Microsoft has said the hackers are likely Chinese. ConnectWise did not respond to TechCrunch's request for comment.
- Security company Barracuda was breached by financially motivated criminals who used an exposed credential belonging to a third-party service whose access had not been revoked after the contract ended.
How the attacks work
The attackers use the tool to scan a victim's network quickly, exploit simple weaknesses such as poor protections and exposed or stolen passwords, and then move through the network to reach databases, mailboxes and other sensitive data, the report said.
The FBI called these intrusions "noisy", meaning they generate heavy traffic that can set off alarms. The difficulty is speed: people may not be able to respond in time.
"Defenders who rely on spotting malicious activity and then responding to it will often arrive after the attacker has finished," Barracuda's chief information security officer Arve Kjoelen told TechCrunch. Barracuda said its logs were not detailed enough to rule out data access, so it chose to "assume the worst" and changed all customer credentials.
The tool is still in circulation
Artex was first released publicly in July as open-source software. Its developer, who is not named in the report, has since made it closed-source, but the FBI said the earlier version remains "widely accessible".
TechCrunch also reported that researchers last week found hundreds of Artex systems reachable on the internet without a password, probably through misconfiguration, including the one used against the South Korean institutions.
What is not known
The report does not say whether the same group was behind the three incidents, which AI models the attackers connected to the tool, or how many Barracuda customers were told to reset their credentials. The FBI advisory itself was not available for independent review, and this account relies on TechCrunch's reporting of it.
The weaknesses described in the incidents were basic ones: missing multi-factor authentication, exposed passwords and old vendor access that was never switched off.
PS · NASA/Pat Izzo




