Cisco Talos warns AI penetration testing frameworks such as ARTEX are being misused in real intrusions
- Severity
- High
- Sources
- 1
What happened
Talos tracks in-the-wild misuse of AI-enabled pentest and red team frameworks (ARTEX, Hephaestus, PentAGI, HexStrike AI) that automate recon, exploitation and follow-on actions using tools such as nmap and nuclei, plus services like Censys, Shodan and residential proxies. ARTEX, published on GitHub in July 2026, is used alongside VShell, GoPhish and AdaptixC2. CrowdStrike attributes an ARTEX-enabled espionage campaign to VAPOR PANDA (over 200 organisations, nine confirmed compromises). Microsoft attributes activity to Storm-0940, which used ConnectWise credentials to target over 2,250 Microsoft 365 tenants. Talos found over 700 reachable PentAGI instances. Defenders should prioritise behavioural detection, since these attacks may lack IOCs, and test containment and incident response plans for machine-speed intrusions.
For security analysts · Written by AI from 1 source · Wrong?
What to do
- Test your incident response plan against fast, automated intrusions that may leave no known indicators of compromise.
- Monitor for behavioural signs of compromise, such as rapid sign-in attempts, credential scanning and unusual privilege changes.
- Review and tighten credential and application access in Microsoft 365 and connected services, given the reported ConnectWise credential compromise.
Written by AI from the sources, checked against them
Full story timeline · 1 update
Who reported · 1 outlet
Who and what is in it
Threat actors
Malware
Companies