AI Cyberattacks Now Move in Hours, FBI and Google Warn

AI Cyberattacks Now Move in Hours, FBI and Google Warn

Three hours is barely enough time for an IT security team to investigate an alert, identify the affected systems, and coordinate a response.

For some cybercriminals, it is now enough time to break into an organization and gain privileged access.

That is the disturbing reality emerging from two major cybersecurity warnings issued in November 2026. The FBI has documented how attackers are using AI-powered automation to compromise enterprise systems at unprecedented speed, while Google says autonomous AI agents are beginning to discover and exploit software vulnerabilities without human intervention.

The concern is no longer whether artificial intelligence will change cybercrime. It is how quickly businesses can adapt now that the change is already happening.

FBI Warning: ARTEX Is Accelerating Cyberattacks

On November 13, 2026, the FBI released a FLASH cybersecurity advisory about ARTEX, an open-source penetration-testing framework that cybercriminals and state-aligned threat actors have turned into an automated attack platform.

ARTEX uses large language models (LLMs) and specialized AI agents to perform activities that previously required considerable human effort.

These include identifying vulnerable systems, testing credentials, exploiting weaknesses, navigating cloud environments, and attempting to steal sensitive information.

The FBI’s findings reveal the scale of the threat:

  • Under 3 hours: Some attacks progressed from initial targeting to privileged access.
  • 2,250+ cloud tenants: Attackers attempted authentication against more than 2,250 Microsoft 365 tenants following the ConnectWise supply-chain compromise.
  • 67,000 individuals: Personal information was reportedly stolen during ARTEX-linked attacks against South Korean financial institutions.
  • 1,600+ instances: Researchers identified more than 1,600 publicly exposed ARTEX consoles as of early November 2026.

The original developer removed ARTEX from GitHub in October. However, copies and modified versions continue circulating online.

Removing the original software did not eliminate the threat.

The ConnectWise Attack Shows What AI Automation Can Do

One of the clearest examples emerged from the October 2026 ConnectWise Cloud Backup supply-chain compromise.

A China-state-aligned group known as Storm-0940 used ARTEX to accelerate reconnaissance, authentication attempts, and abuse of cloud application permissions.

During a related password-spraying campaign, Microsoft recorded more than 151,000 failed sign-ins, including a peak of approximately 2,900 attempts per minute.

That particular password-spraying activity produced no observed successful authentication. However, separate activity involving compromised ConnectWise application identities enabled unauthorized authentication and follow-on operations in multiple customer environments.

Importantly, the 2,250-plus tenants represented attempted access, not 2,250 confirmed breaches.

The incident exposed a weakness that many enterprises overlook: attackers do not always need to compromise an employee account when a trusted third-party application already has powerful permissions.

A compromised integration can become an entry point into systems that otherwise appear securely configured.

Google Warns That AI Is Learning to Exploit Patches

Three days after the FBI advisory, the Google Threat Intelligence Group published research describing another concerning development.

Attackers are using autonomous AI frameworks to monitor software updates, compare code changes, identify the vulnerabilities those updates address, and develop exploits.

Some of these workflows operate without a human directing each step.

Traditionally, analyzing a security patch and turning the findings into a working exploit required significant expertise and manual effort.

AI can dramatically reduce that workload.

Google also documented the use of agentic penetration-testing tools against real organizations, including activity involving the US aerospace sector.

The underlying attack methods are not necessarily new. Criminals still exploit weak credentials, excessive permissions, exposed applications, and unpatched software.

What has changed is how quickly those weaknesses can be discovered and exploited.

What This Means for Enterprise Security

Consider an enterprise operating hundreds of applications across cloud environments, employee devices, external vendors, and internal business systems.

Its primary infrastructure may be carefully monitored, but an older supplier portal, forgotten API, or overprivileged integration might receive considerably less attention.

Those overlooked systems are precisely the kinds of entry points automated attackers can investigate at scale.

Three enterprise security risks deserve particular attention.

The shrinking response window: A security alert that might once have allowed several hours for investigation could now require action within minutes.

Third-party access exposure: Connected applications, service accounts, and external vendors can create paths into sensitive business environments even when employee authentication is strong.

Faster vulnerability exploitation: Traditional patch schedules may struggle when attackers can rapidly analyze newly released updates and target organizations that have not yet applied them.

For enterprises, the consequences extend beyond cybersecurity incidents. Compromised systems can interrupt operations, expose customer data, trigger compliance investigations, and undermine business relationships.

How Should Enterprises Respond to AI-Powered Cyberattacks?

The FBI and Google findings point toward a practical priority: reduce the opportunities attackers can exploit while improving detection and response speed.

Organizations should prioritize five measures:

  1. Review cloud identities and permissions. Audit privileged accounts, service principals, third-party integrations, and unnecessary application access.
  2. Strengthen browser and endpoint security. Apply appropriate access policies and improve visibility into applications, devices, and risky user activity.
  3. Continuously assess exposed systems. Include legacy applications, APIs, supplier portals, and cloud management interfaces.
  4. Improve automated detection. Monitor unusual authentication patterns, rapid API requests, unexpected permission changes, and suspicious data transfers.
  5. Test incident-response readiness. Establish clear containment procedures and regularly evaluate whether teams can respond to rapidly developing attacks.

Google is also advancing defensive AI capabilities through Google AI Threat Defense, combining threat intelligence, AI-assisted vulnerability discovery, automated remediation technologies, and security expertise.

As a Google Cloud Partner and Chrome Service Partner, Codimite works across enterprise browser security, endpoint modernization, cloud infrastructure, and secure AI integration. These areas are becoming increasingly interconnected as organizations rethink how access, applications, and sensitive data should be protected.

The objective is not simply to introduce more security tools. It is to build environments where risks are visible, access is controlled, and security measures can keep pace with increasingly automated threats.

The Next Cybersecurity Challenge Is Speed

The FBI and Google’s warnings point to a fundamental change.

AI is not inventing every attack technique from scratch. It is making established techniques faster, easier to coordinate, and possible to execute across more targets simultaneously.

Enterprises that rely entirely on manual monitoring and delayed responses may find themselves struggling to keep up.

The question for business leaders is becoming increasingly straightforward:

If an attacker can gain privileged access in under three hours, how quickly can your organization detect and stop them?

We value your privacy

Codimite uses essential cookies to keep our website secure and functional. With your consent, we also use analytics and marketing cookies to improve your experience and understand website usage.

You can accept all cookies, reject all cookies, or manage your preferences. Learn more in our Privacy Policy.

We use cookies to understand how our website is used. You can or . See our Privacy Policy.