KELA：2026年年中人工智能威胁态势报告（英文版）（44页）.pdf

1、 2026 Mid-Year AI Threat Landscape Report July 2026 Table of contents Executive Summary.3 The Agentic Shift and Machine-Speed Exploitation.5 Introduction:The Commoditization of the Agentic Threat.5 Technical Foundations of Autonomous Vulnerability Discovery&Exploitation(AVDE).9 Agentic Cyber Espiona

2、ge:Subversion and Autonomous Execution.10 AI-Driven Execution Across the Attack Chain.11 The Dual Threat of AI Manipulation:Vibe Hacking vs.AI-Enhanced Social Engineering.12 Attacking the AI:The Rise of“Vibe Hacking”.13 The Insider AI Threat(Inter-Agent Trust Exploitation).20 The Infostealer Evoluti

3、on:Harvesting the Agentic Plaintext Layer.22 Threat Actor Chatter and Observed Cybercrime Activity.23 AI Session Hijacking:The High-Leverage Entry Point.24 Cognitive Context Theft-The Rise of Local AI Memories.30 Darknet Data Lake Insights:Tracking the Autonomous Underground.39 Active Reconnaissance

4、&Surface Mapping.39 Enterprise Defense:Hardening for the Agentic Era.43 The Imperative of CTI-Driven Posture Hardening.43 Threat Detection:Identifying the Tells of Agentic Execution.43 Strategic Hardening&Architectural Containment.44 Mid-Year AI Threat Landscape Report-2026|2 Executive Summary The C

5、ommoditization of the Agentic Threat The year 2026 marks a structural inflection point in the cyber threat landscape,driven by the rapid commoditization of open-source agentic AI within the cybercriminal underground.While proprietary frontier models initiated this paradigm shift-most notably Anthrop

6、ics Claude Mythos,which pioneered Autonomous Vulnerability Discovery and Exploitation(AVDE),and its equivalent,OpenAIs GPT-5.5-Cyber-the immediate,operational threat originates from open-source alternatives.Threat actors are aggressively adopting decentralized,locally hosted models(such as Qwen,Deep