Reimagine Developer Engagement
with Doc-E.ai

January 17, 2026

Unexpected Code Execution (ASI05) is a critical vulnerability where an AI agent translates natural-language instructions into executable code (Python, Bash, JavaScript) that is then run in an insecure environment. Unlike traditional Remote Code Execution (RCE), ASI05 occurs when an agent’s "vibe-coding" or autonomous loop is tricked into generating and executing its own malicious payload. The AutoGPT RCE vulnerabilities serve as the primary case study for this "Self-Pwn" attack vector.

January 16, 2026

Agentic Supply Chain Vulnerability (ASI04) refers to the risk of malicious code or instructions being introduced into an agent’s ecosystem via third-party components, such as Model Context Protocol (MCP) servers, tool definitions, or poisoned data connectors. As agents move toward dynamic, "just-in-time" tool loading, the supply chain becomes the primary vector for silent environment compromise. The 2025 GitHub MCP Exploit highlighted how a single poisoned repository could hijack local developer environments.

January 15, 2026

Identity & Privilege Abuse (ASI03) is a high-impact security risk where AI agents operate with excessive, static, or inherited permissions that allow them to act beyond their intended scope. This often leads to Credential Leaking or the "Confused Deputy" attack, where one agent misuses the privileges of another. Securing ASI03 requires a shift from human-centric IAM to Agent-Specific Identity and ephemeral, short-lived tokens.

January 14, 2026

Tool Misuse (ASI02) occurs when an AI agent utilizes its authorized toolset (APIs, CLI, Database access) to perform unintended, destructive, or unauthorized actions. This risk is primarily driven by Excessive Agency, where the agent's permissions exceed the requirements of its specific task. The 2025 Amazon Q incident serves as the benchmark case study for how natural language can be manipulated to trigger "wiper-style" destructive commands.

January 13, 2026

Agent Goal Hijack (ASI01) is a critical security vulnerability where an autonomous agent's primary objective is displaced by adversarial instructions embedded in data. Unlike standard prompt injection, goal hijacking persists through the agent’s planning phase. The EchoLeak exploit is the primary 2026 example, where agents were tricked into silent data exfiltration via natural language redirection.

January 12, 2026

The OWASP Top 10 for Agentic Applications (2026) is the industry-standard framework for securing autonomous AI. Unlike traditional LLM risks (like prompt injection), Agentic Security (ASI) focuses on the risks of AI that can plan, use tools, and interact with other agents. Key vulnerabilities include Agent Goal Hijacking (ASI01), Tool Misuse (ASI02), and Agentic Supply Chain Vulnerabilities (ASI04).