Cisco Talos Reveals How Hackers Use ARTEX, PentAGI, and HexStrike AI to Automate Cyberattacks

Blog WriterCybersecurity News - Original News Source is cybersecuritynews.com


AI-powered penetration testing tools are helping hackers automate reconnaissance, vulnerability testing, and follow-on attacks with less hands-on work.

Cisco Talos warns that these platforms can connect familiar security utilities with AI agents, allowing operators to pursue targets faster and run several tasks at once.

The danger comes from how legitimate testing capabilities are being used against organizations without permission. ARTEX, PentAGI, and HexStrike AI illustrate different approaches to this automation.

Instead of simply suggesting commands, these frameworks connect AI models with tools that perform security tasks. Their growing availability lowers technical barriers for attackers, while giving experienced operators ways to scale operations and shorten the time defenders have to respond.

Cisco Talos researchers identified growing malicious use of these frameworks in a November 13 report, including ARTEX activity associated with credential theft and malware deployment.

The distinction matters: these platforms are dual-use security tools, not malware families themselves. Earlier reporting on ARTEX AI attack automation explains how agents turn operator goals into connected technical actions.

Hackers Use ARTEX, PentAGI, and HexStrike AI

ARTEX appeared on GitHub in July 2026 under the developer name Autumn-27. An operator supplies a target and objective, and a planner agent divides the work into tasks.

Worker agents execute those tasks using security tools, while a shared graph records discovered assets, findings, and progress. This structure lets subsequent actions build on earlier results with relatively little operator input.

PentAGI provides a self-hosted, multi-agent testing environment with a web interface for monitoring activity and viewing attack-chain graphs. Talos found more than 700 reachable PentAGI instances online.

HexStrike AI connects with Claude Code, VS Code, and other clients supporting Model Context Protocol, or MCP. These differences concern how operators connect, manage, and observe automated workflows, rather than proving equivalent malicious activity across all three platforms.

ARTEX has the clearest campaign evidence in the report. Talos cites Microsoft findings that Storm-0940 used compromised ConnectWise application credentials to accelerate sign-in attempts against more than 2,250 downstream Microsoft 365 tenants.

The previously documented Microsoft 365 tenant attacks show how trusted application access can extend a compromise into connected customer environments.

Other reported activity involved password spraying, weak administrative credentials, missing API authentication, command injection, and authorization flaws.

Agents also searched public repositories for exposed secrets. In one Microsoft-observed case, the transition from GitHub searches to Azure Key Vault data theft took five minutes. Reporting on GitHub secrets theft automation details how leaked credentials became a bridge into cloud resources.

Detection and Containment Must Match Automated Attack Speeds

Talos identified more than 2,500 ARTEX-associated infrastructure IP addresses and examined tools present during overlapping periods. These included PentAGI, VShell, AdaptixC2, Mimikatz, Cobalt Strike, and several scanners.

However, shared infrastructure does not prove that ARTEX launched every tool or that one operator controlled everything. ARTEX also does not ship with a traditional command-and-control framework, although MCP allows operators to add compatible servers.

That distinction matters when assessing alerts. A recognizable tool or user agent can provide an investigative lead, but defenders need supporting evidence.

Talos recommends correlating authentication, endpoint, and cloud API logs for rapid requests, unfamiliar applications, unusual token use, and privilege changes. Its findings reinforce the concerns described in the FBI ARTEX security warning about attackers compressing intrusion timelines.

Organizations should patch exposed systems, remove default passwords, enforce multifactor authentication, and apply rate limits against credential attacks.

Public repositories need secret scanning, while exposed credentials require immediate revocation and rotation. Least-privilege application permissions can reduce downstream access.

Talos also advises endpoint tamper protection and tested procedures for disabling suspicious applications, revoking tokens, and isolating affected systems quickly.

Indicators of compromise (IoCs):-

Indicator type Source-listed value Context
IPv4 92.255.57.95 ARTEX-associated address
IPv4 38.38.174.152 ARTEX-associated address
IPv4 62.171.139.26 ARTEX-associated address
IPv4 155.2.192.70 ARTEX-associated address
IPv4 38.12.90.84 ARTEX-associated address

Note: IP addresses and domains are intentionally Qdefanged (e.g., [.]) to prevent accidental resolution or hyperlinking. Re-fang only within controlled threat intelligence platforms such as MISP, VirusTotal, or your SIEM.

Give your SOC the alert context on the latest threats to stop incidents quickly -> Integrate Threat Intel from 16K SOCs

The post Cisco Talos Reveals How Hackers Use ARTEX, PentAGI, and HexStrike AI to Automate Cyberattacks appeared first on Cyber Security News.