Talos finds 700+ internet-exposed PentAGI AI pentest instances as ARTEX framework fuels VAPOR PANDA and Storm-0940 intrusions
Cisco Talos identified more than 700 internet-reachable PentAGI instances through a Censys search on the page title. The count covers reachable deployments only. It does not show malicious use, compromise or unauthenticated access. PentAGI is an open-source, self-hosted platform that coordinates AI agents to run security tools. Talos describes a wider ecosystem of AI-orchestrated offensive frameworks, including ARTEX (published in July 2026 under the pseudonym Autumn-27) and the earlier Hephaestus toolkit. Talos reports ARTEX adoption by financially motivated criminals and China-nexus operators. This record rests on the article alone. No vendor or primary-source research was available, so the details have not been checked against primary sources.
Detalhes técnicos
- CrowdStrike linked VAPOR PANDA to ARTEX activity against more than 200 organizations in at least 14 countries, starting October 17, 2026. Post-exploitation was confirmed at nine organizations. Operators kept control at decision points while agents pursued parallel paths across applications, cloud, identity and remote access.
- Microsoft attributed ARTEX-driven activity to Storm-0940 after ConnectWise application credentials were compromised. The actor attempted authentication against more than 2,250 downstream tenants. Some tenants showed Microsoft Graph reconnaissance, credential modifications and privilege changes. Successful authentication alone did not prove data access or exfiltration.
- A separate Microsoft investigation documented an agent going from GitHub credential discovery to Azure Key Vault exfiltration in five minutes.
- Talos identified more than 2,500 ARTEX-associated C2 IP addresses. Observed tools overlapping with them included PentAGI, VShell, AdaptixC2, Mimikatz and Cobalt Strike. This co-occurrence does not prove that ARTEX used every tool or that one actor controlled all the activity.
- Talos found that operator expertise strongly shaped outcomes. AI amplified existing skills rather than removing technical limits.
Remediação
- Correlate authentication, endpoint and cloud telemetry for rapid retries, unusual token use, application credential additions and privilege changes.
- Revoke exposed secrets and limit workload identity privileges.
- Investigate behavior rather than changeable IP addresses or user agents.
- Prepare containment procedures, preserve evidence, disable confirmed malicious access and check for persistence across connected systems.
Linha do tempo
-
Talos finds 700+ internet-exposed PentAGI AI pentest instances as ARTEX framework fuels VAPOR PANDA and Storm-0940 intrusions
ler a análise desta fonte ocultar
Cisco Talos identified more than 700 internet-reachable PentAGI instances through a Censys search on the page title. The count covers reachable deployments only. It does not show malicious use, compromise or unauthenticated access. PentAGI is an open-source, self-hosted platform that coordinates AI agents to run security tools. Talos describes a wider ecosystem of AI-orchestrated offensive frameworks, including ARTEX (published in July 2026 under the pseudonym Autumn-27) and the earlier Hephaestus toolkit. Talos reports ARTEX adoption by financially motivated criminals and China-nexus operators. This record rests on the article alone. No vendor or primary-source research was available, so the details have not been checked against primary sources.
Technical Details
- CrowdStrike linked VAPOR PANDA to ARTEX activity against more than 200 organizations in at least 14 countries, starting October 17, 2026. Post-exploitation was confirmed at nine organizations. Operators kept control at decision points while agents pursued parallel paths across applications, cloud, identity and remote access.
- Microsoft attributed ARTEX-driven activity to Storm-0940 after ConnectWise application credentials were compromised. The actor attempted authentication against more than 2,250 downstream tenants. Some tenants showed Microsoft Graph reconnaissance, credential modifications and privilege changes. Successful authentication alone did not prove data access or exfiltration.
- A separate Microsoft investigation documented an agent going from GitHub credential discovery to Azure Key Vault exfiltration in five minutes.
- Talos identified more than 2,500 ARTEX-associated C2 IP addresses. Observed tools overlapping with them included PentAGI, VShell, AdaptixC2, Mimikatz and Cobalt Strike. This co-occurrence does not prove that ARTEX used every tool or that one actor controlled all the activity.
- Talos found that operator expertise strongly shaped outcomes. AI amplified existing skills rather than removing technical limits.
Remediation
- Correlate authentication, endpoint and cloud telemetry for rapid retries, unusual token use, application credential additions and privilege changes.
- Revoke exposed secrets and limit workload identity privileges.
- Investigate behavior rather than changeable IP addresses or user agents.
- Prepare containment procedures, preserve evidence, disable confirmed malicious access and check for persistence across connected systems.