Threat

FBI warns ARTEX AI hacking tool is powering global cyberattacks

Page published · Page updated

Dossier

Earliest dated coverage: 15 Nov 2026 · First observed: 15 Nov 2026 · Latest dated coverage: 15 Nov 2026

Coverage timeline

15 Nov 2026bleepingcomputer.com

Single-source incident — one report is available.

Why it matters

ARTEX-driven intrusions reportedly progress from initial targeting to privileged access in under three hours, sharply reducing defenders’ response window while enabling simultaneous attacks against multiple organizations.

The FBI warns that threat actors are using ARTEX, an autonomous AI-agent penetration-testing framework, to automate intrusions worldwide, including a reported ConnectWise supply-chain breach affecting Microsoft 365 customer environments. ARTEX uses external LLMs to coordinate reconnaissance, exploitation, privilege escalation, and data theft with little human intervention.

campaign

Summary

BleepingComputer reports that the FBI is warning of widespread malicious use of ARTEX, an autonomous penetration-testing framework, by multiple threat actors since at least August 2026. The reported activity spans financial, government, technology, defense, healthcare, and other sectors. The principal operational concern is speed and parallelism: the advisory is quoted as describing some intrusions that progressed from initial targeting to privileged access in under three hours.[0]

The most concrete named-actor case is the October ConnectWise Cloud Backup supply-chain campaign, which the article says Microsoft attributed to China-state-aligned Storm-0940. Attackers compromised application identities used by the backup service and authenticated to downstream Microsoft 365 environments using previously granted application permissions. The report distinguishes attempts against more than 2,250 tenants from Huntress's identification of at least 162 compromised tenants. Some intrusions reportedly advanced to credential modifications and persistence attempts.[0]

ARTEX use is not confined to that campaign. The article also describes a Chinese-speaking actor targeting South Korean banks, a separate South Korean criminal investigation, and a Vapor Panda espionage campaign targeting more than 200 organizations across 14 countries. These are distinct reported operations sharing tooling, not evidence that all ARTEX activity belongs to one actor.[0]

The evidence supplied for this dossier is a single news article summarizing findings attributed to the FBI, Microsoft, Huntress, CrowdStrike, and MalwareMustDie. Their underlying reports are not supplied as evidence bodies, so attribution, impact figures, and performance claims remain single-source reporting here. The article provides relevant defensive recommendations but no literal campaign-specific indicators or affected software version ranges.[0]

Attack chain

  1. 1. Compromise of upstream application identities: In the reported ConnectWise campaign, attackers compromised application identities used by ConnectWise Cloud Backup, formerly SkyKick Cloud Backup. The article does not explain how those identities were initially acquired, so it does not support assigning a specific exploit or credential-acquisition method to this step.[0]
  2. 2. Authentication to downstream customer tenants: After compromising those identities, attackers used ARTEX to attempt authentication against more than 2,250 Microsoft 365 tenants. Access relied on permissions customers had previously granted to the backup applications. The article separately reports at least 162 compromised tenants identified by Huntress.[0]
  3. 3. Credential modification and persistence attempts: Some downstream intrusions progressed to credential modifications and attempts to establish persistent access. The article does not establish that these actions succeeded in every compromised tenant or describe a subsequent data-exfiltration stage for this campaign.[0]

Disclosure timeline

DateEvent
2026-07According to the article, developer Autumn-27 originally released ARTEX on GitHub.[0]
2026-08The FBI's reported observation window for malicious ARTEX use began no later than August 2026.[0]
2026-10-08The original developer reportedly removed the ARTEX repository after reports of threat-actor abuse. Forks, archives, localized versions, and a commercial edition reportedly remained available.[0]
2026-10Microsoft reportedly published its investigation attributing the ConnectWise Cloud Backup campaign to Storm-0940 and describing ARTEX-assisted reconnaissance, password spraying, and abuse of cloud application permissions.[0]
Early November 2026Researchers reportedly had identified more than 1,600 publicly exposed ARTEX consoles, with the actual deployment count believed to be higher.[0]
2026-11-15BleepingComputer published the supplied article, reporting that the FBI FLASH advisory and MalwareMustDie detection-signature pack had been released the preceding Friday.[0]

Actor profile

Storm-0940

The article reports Microsoft's attribution of the October ConnectWise Cloud Backup campaign to Storm-0940, described as China-state-aligned. Reported ARTEX-assisted activities included reconnaissance, password spraying, and abuse of cloud application permissions. The campaign subsequently attempted authentication to more than 2,250 downstream Microsoft 365 tenants. This attribution applies to the ConnectWise operation; the article separately describes other ARTEX users and does not establish common control over them.[0]

How it works

The concrete access mechanism described for the ConnectWise campaign is abuse of trusted application identities rather than a named software vulnerability. Compromising identities used by the cloud-backup service allowed attackers to authenticate to customer tenants with permissions already granted to those applications. The article does not identify a CVE, an initial exploit, or a software version boundary for that compromise.[0]

ARTEX supplies an automation layer around intrusion activities. Operators provide goals and tasks, which the Go-based framework translates into commands executed by specialized AI agents. Rather than operating its own model, it integrates external LLMs, including Claude, GPT, DeepSeek, GLM, Grok, and Kimi. Reported capabilities include reconnaissance, vulnerability discovery, exploitation, privilege escalation, and data theft with limited human intervention.[0]

The reported advantage is accelerated exploration and chaining of attack paths, including against partner portals, employee applications, legacy services, and third-party integrations. In the separate Vapor Panda operation, agents reportedly pursued alternative paths after initial attempts failed. These capabilities should not be read as proof that every listed technique occurred in the ConnectWise campaign.[0]

Key takeaways

  • Treat ARTEX as shared tooling, not an attribution marker. The article describes its use by Storm-0940, Vapor Panda, a Chinese-speaking banking attacker, and a separately investigated criminal group.[0]
  • The strongest documented campaign mechanism is downstream abuse of trusted cloud application identities. Authentication attempts against more than 2,250 tenants and at least 162 reported compromises must remain distinct scope measurements.[0]
  • Repository removal did not reportedly eliminate availability: forks, archives, localized versions, and ARTEX PRO remained accessible. The reported exposed-console count measures visible tooling deployments, not a verified count of malicious operators or victims.[0]
  • This evidence supports behavior-focused monitoring and cloud-identity review, but not a version-specific patch directive or an extracted IOC blocklist. No concrete malicious IP address, domain, file hash, or CVE is supplied in the article.[0]

Defensive actions

  • Secure internet-facing applications, including partner portals, employee applications, legacy services, and third-party integrations.: The reported FBI assessment identifies overlooked systems in these categories as particularly useful to ARTEX-driven vulnerability discovery. Securing internet-facing applications is an explicit recommendation in the article.[0]
  • Restrict access to cloud administration interfaces.: The article explicitly recommends this control alongside its account of ARTEX-enabled cloud intrusions and abuse of application permissions.[0]
  • Review cloud application permissions and authentication logs, with attention to third-party backup application identities.: The ConnectWise campaign reportedly used compromised backup application identities and existing customer-granted permissions to reach Microsoft 365 tenants. Permission and authentication-log review directly addresses that documented access mechanism.[0]
  • Implement rate limits on sensitive APIs and monitor unusual bursts of automated requests.: Both measures are recommended in the article. ARTEX can generate substantial suspicious traffic while automating concurrent intrusion tasks, but the reported speed of some attacks leaves a short response window.[0]
  • Monitor suspicious cloud activity, unexpected credential or privilege changes, and large data transfers.: These are explicit monitoring recommendations in the report. Credential-change monitoring is especially pertinent to the ConnectWise case, in which some intrusions reportedly progressed to credential modifications and persistence attempts.[0]
  • Evaluate the reported ARTEX detection-signature pack and playbook, applying their confidence distinctions and validating infrastructure indicators before operational use.: The article describes coverage for network requests, identity abuse, cloud activity, malicious files, and automated attack patterns. It also warns that IP addresses and other infrastructure indicators can become stale or produce false positives; the supplied article does not include the underlying rules or indicator values.[0]
  • Report identified ARTEX-associated suspicious activity to the FBI or IC3.: The article explicitly relays the FBI's request for organizations identifying associated suspicious activity to submit their findings.[0]