Safety & Security

FBI Warns Hackers Use AI Tool Artex to Break In Within 3 Hours

The FBI says attackers using the AI tool Artex went from target selection to high-level access in under 3 hours, hitting banks, governments and hospitals since August.

Hackers Are Using AI to Speed Through Cyberattacks, FBI Warns
Hackers Are Using AI to Speed Through Cyberattacks, FBI Warnsfreephotocc / pixabay
By Rebecca Stone5 min read

Why it matters

  • The FBI said some Artex-assisted attacks took less than 3 hours from target selection to high-level access.
  • Attackers have used Artex since at least August against banks, government agencies, tech companies and healthcare providers.
  • Microsoft says attackers breached over 2,250 organizations' Microsoft 365 systems using stolen ConnectWise backup credentials alongside Artex.
  • The FBI cited attacks on South Korean financial institutions that exposed personal data of roughly 67,000 people.
  • The developer removed the original Artex GitHub repository in October, but copies remain available.

Some cyberattacks now run from target selection to high-level system access in under 3 hours, and the FBI attributes that speed to an AI-powered security testing tool called Artex. In an advisory issued Friday, the agency said hackers have used the tool since at least August against banks, government agencies, technology companies and healthcare providers worldwide.

The FBI published the warning the same day communications problems forced a nationwide Federal Aviation Administration ground stop, and one day after a major Microsoft 365 outage began. Microsoft attributed its disruption to technical problems. The FAA said it found no indication of a cyberattack. An FBI official told ABC News that Artex does not appear to be connected to either incident, or to other recent disruptions, including banking outages involving JPMorgan Chase.

The advisory lands amid growing concern about the security risks of AI agents, which can carry out tasks with little human supervision. Artex shows how that capability cuts both ways: a tool built to find security flaws before criminals do can, in the wrong hands, compress an intrusion campaign into a single afternoon.

What is Artex and how do attackers use it?

Artex is an open-source AI-powered security testing tool. It automates the security tests organizations run to uncover their own weaknesses. According to Microsoft's analysis, its agents can examine defenses, exploit flaws and extract data — the same chain of steps an offensive hacker would follow manually.

That automation changes the economics of an attack. A human intruder probes one entry point at a time. Attackers using Artex can try several ways into a network simultaneously, the FBI said. The automated attacks can trigger security alerts, but their speed leaves defenders little time to respond. A three-hour window from target selection to privileged access leaves almost no margin for human review of those alerts.

How many victims has the campaign claimed?

The FBI cited attacks against South Korean financial institutions that exposed the personal information of approximately 67,000 people. That figure covers a single country and a single sector, which suggests the full scope of the campaign may be larger.

Microsoft documented a separate and larger toll. In its own campaign analysis, Microsoft warned that attackers breached the Microsoft 365 systems of more than 2,250 organizations. In those intrusions, the attackers paired Artex with stolen ConnectWise backup credentials.

The credential abuse carried a compounding effect. Customers had already authorized the backup software, so attackers could use its access without logging in as an individual user. Stolen third-party access bypassed the user-level authentication that defenders traditionally watch. Microsoft added one caution: a successful sign-in does not necessarily mean data was accessed or stolen.

What did Amazon's researchers find?

Amazon Web Services published research on Nov. 13 describing a related scheme with a distinctive twist: stealing credentials for AI services, then using that hijacked access to hunt for more credentials at someone else's expense. The victims of the initial theft effectively paid the compute bill for the attackers' next round of discovery.

Amazon Threat Intelligence uncovered the activity through decoy systems posing as AI services — honeypots built to look like the kind of AI endpoints credential thieves target. The company linked the campaigns to Artex with moderate confidence. It could not independently verify the results the agents reported on systems outside its monitoring.

Amazon's findings point to the enabling conditions. The attacks relied on default passwords, exposed login details and outdated software — long-standing failures that hackers can now find and exploit faster with AI. The vulnerability is not new. The speed of weaponizing it is.

Where did the tool come from and can it be stopped?

Artex was originally published as an open-source project on GitHub. Its developer removed the original repository in October after reports of abuse, according to the FBI. Copies remain available. That trajectory — security research tool, criminal adoption, takedown, redistribution — matches the pattern of dual-use offensive security software, and the FBI's advisory makes clear the removal did not end the threat.

The agency recommends four defensive measures:

  • Limit access to sensitive systems.
  • Review software permissions.
  • Replace exposed credentials.
  • Monitor for unusual account activity or large data transfers.

Each measure addresses a specific weakness the campaign exploited. Access limits constrain what a hijacked agent can reach. Permission reviews catch the pre-authorized third-party access, like ConnectWise backups, that let attackers skip user logins. Credential replacement closes the door on stolen keys. Transfer monitoring targets the data-extraction stage, the point where a breach becomes a loss.

Why does this matter beyond the immediate victims?

The story matters because it marks a measurable shift in attack speed, not just a new tool. A sub-3-hour intrusion timeline, verified breaches at more than 2,250 organizations and 67,000 compromised individuals in South Korea give defenders concrete numbers for what AI-accelerated offense looks like in practice. Microsoft's analysis and Amazon's Nov. 13 research independently describe the same pattern: automation that probes, exploits and extracts faster than alert-and-response cycles can absorb.

The dual-use problem compounds it. Artex exists because organizations need automated security testing, and its removal from GitHub did not remove it from circulation. As long as open-source agents capable of examining defenses, exploiting flaws and extracting data remain copyable, defenders will face the same tools they use, pointed the other way — and the FBI's timeline suggests the response window will keep shrinking.

Source: cnet.com

Share this article:

More from Rebecca Stone

Rebecca Stone

Show full bio

Correspondent covering companies, deals and strategy in artificial intelligence and machine learning at AI In Context.

728 articles

Related articles

  1. FBI Warns ARTEX AI Agents Compress Cyberattacks Into Hours
  2. CrowdStrike: AI Hacking Tool Let One Attacker Breach Korean Banks
  3. Partnership on AI: Agent Frameworks Can't Show What Agents Did
  4. One Israeli Startup Sits Behind a String of Rogue AI Disclosures
  5. One Hacker, a Few Thousand Dollars, and an AI Army That Breached John Deere

« Previous articleNext article »