----------------
🎯 AI
===================
The FBI says threat actors are using Artex, an open-source AI-assisted penetration testing framework, to automate parts of intrusions from reconnaissance through data exfiltration. Some operations progressed from initial targeting to privileged access in under three hours, according to the agency.
🔹 Framework and observations
Artex translates high-level objectives written in natural language into commands executed by AI agents. The FBI describes operations requiring limited human interaction after objectives are set, including concurrent activity against multiple targets. The source does not specify the commands used in the reported intrusions or establish that every stage was fully autonomous.
The framework can integrate models including Claude, GPT, DeepSeek, GLM, Grok and Kimi. The FBI says its analysis of attacker-controlled instances found Chinese models were most common, with DeepSeek Flash models dominating the observed usage. The agency attributes that preference to capability, relatively low cost and a lack of observable guardrails when requests were framed as penetration testing.
Research cited by the FBI found more than 1,600 publicly exposed Artex consoles. That is a count of exposed instances, not confirmed malicious installations. Censys estimates total deployments are likely higher when private networks and systems behind firewalls are considered.
🔹 Attack Chain Analysis
The source describes framework capabilities and campaign observations rather than one complete intrusion chain:
• 🔍 Reconnaissance: Artex agents can automate information gathering against targets.
• 🛡️ Privileged access: The FBI reports that some intrusions reached this point in under three hours from initial targeting. The access and escalation methods are not specified.
• 📤 Data exfiltration: The framework can automate this phase, according to the FBI. The source does not identify the destination or transport used.
Amazon separately describes Chinese-nexus groups stealing credentials used to access AI models. In that operation, stolen keys paid for model inference that supported further credential hunting. Amazon also reports a separate campaign against cryptocurrency exchanges involving real-time deepfake injection to bypass know-your-customer verification. The supplied text does not detail that campaign's initial access or injection implementation.
🔹 Reported incidents and limits
The article associates Artex with attacks on South Korean and Japanese financial institutions. Microsoft also attributes a supply-chain attack on IT management software company ConnectWise involving Artex to Chinese state-backed actors. The FBI lists targets across government, finance, technology, defense, semiconductors, energy, healthcare, manufacturing, telecommunications and legal services.
The original GitHub repository was removed after developer Li Puhua said malicious actors had abused the framework. The FBI says forks, localized variants, archived copies and a commercial version remain available.
The supplied article ends mid-sentence and contains no indicators, detection rules, CVEs or specific mitigation guidance. It does not provide enough detail to reproduce the reported intrusion paths or independently verify the console count.
🔹 Artex #AI #ThreatIntelligence #tool
🔗 Source: https://therecord.media/artex-ai-tool-fbi-flash-alert