PT-2026-50689

Name of the Vulnerable Software and Affected Versions Eclipse Theia versions prior to 1.71.0 Description The AI chat agent processes workspace file and directory names as part of its prompt context without distinguishing them from system instructions. This allows for indirect prompt injection,...

Hiding in Plain Floats: Steganographic Carriers for Indirect Prompt and Content Injection

Text-centered prompt-injection defenses assume that the malicious signal is visible in one of the inspected text views. We study a reproducible LLM01-style indirect prompt/content-injection failure mode where that assumption breaks: a payload caught in plain English slips past the same detector...

AgentRedBench: Dynamic Redteaming and Integration-Aware Defense for LLM Agents over SaaS Integrations

Indirect prompt injection in tool-use agents is a concrete production threat: LLM agents read from integrations third-party services such as Gmail, Salesforce, or Jira accessed through tool calls whose response content the user neither writes nor controls. Existing benchmarks under-measure the...

IPI-Proxy: An Intercepting Proxy for Red-Teaming Web-Browsing AI Agents against Indirect Prompt Injection

Web-browsing AI agents are increasingly deployed in enterprise settings under strict whitelists of approved domains, yet adversaries can still influence them by embedding hidden instructions in the HTML pages those domains serve. Existing red-teaming resources fall short of this scenario:...

Indirect Prompt Injection in the Wild: An Empirical Study of Prevalence, Techniques, and Objectives

As LLMs are increasingly integrated into systems that browse, retrieve, summarize, and act on web content, webpages have become an untrusted input vector for downstream model behavior. This enables site owners, contributors, and adversaries to embed instructions directly in web resources, i.e.,...

Logic-to-Code Execution via Indirect Prompt Injection

This document explores a critical architectural vulnerability in Large Language Model LLM implementations, specifically within Command Line Interface CLI tools and automated agentic workflows. The research demonstrates how the absence of separation between the control plane instructions and the...

Hackers Use Hidden Website Instructions in New Attacks on AI Assistants

Cybersecurity researchers at Forcepoint uncover new indirect prompt injection attacks that use hidden website code to exploit AI assistants like GitHub Copilot...

GrafanaGhost Vulnerability Allows Data Theft via AI Injection

GrafanaGhost is a critical vulnerability in Grafana’s AI components that uses indirect prompt injection and protocol-relative URL bypasses to exfiltrate data...

CVE-2026-33654

nanobot is a personal AI assistant. Prior to version 0.1.6, an indirect prompt injection vulnerability exists in the email channel processing module nanobot/channels/email.py, allowing a remote, unauthenticated attacker to execute arbitrary LLM instructions and subsequently, system tools without...

CVE-2026-33654 Zero-Click Indirect Prompt Injection and Authentication Bypass via Email Polling

nanobot is a personal AI assistant. Prior to version 0.1.6, an indirect prompt injection vulnerability exists in the email channel processing module nanobot/channels/email.py, allowing a remote, unauthenticated attacker to execute arbitrary LLM instructions and subsequently, system tools without...

CVE-2026-33654

nanobot is a personal AI assistant. Prior to version 0.1.6, an indirect prompt injection vulnerability exists in the email channel processing module nanobot/channels/email.py, allowing a remote, unauthenticated attacker to execute arbitrary LLM instructions and subsequently, system tools without...

OpenClaw AI Agent Flaws Could Enable Prompt Injection and Data Exfiltration

China's National Computer Network Emergency Response Technical Team CNCERT has issued a warning about the security risks stemming from the use of OpenClaw formerly Clawdbot and Moltbot, an open-source and self-hosted autonomous artificial intelligence AI agent. In a post shared on WeChat, CNCERT...

Security Considerations for Artificial Intelligence Agents

This article, a lightly adapted version of Perplexity's response to NIST/CAISI Request for Information 2025-0035, details our observations and recommendations concerning the security of frontier AI agents. These insights are informed by Perplexity's experience operating general-purpose agentic...

AttriGuard: Defeating Indirect Prompt Injection in LLM Agents Via Causal Attribution of Tool Invocations

LLM agents are highly vulnerable to Indirect Prompt Injection IPI, where adversaries embed malicious directives in untrusted tool outputs to hijack execution. Most existing defenses treat IPI as an input-level semantic discrimination problem, which often fails to generalize to unseen payloads. We...

CVE-2026-30856

WeKnora is an LLM-powered framework designed for deep document understanding and semantic retrieval. Prior to version 0.3.0, a vulnerability involving tool name collision and indirect prompt injection allows a malicious remote MCP server to hijack tool execution. By exploiting an ambiguous naming...

CVE-2026-30856

WeKnora CVE-2026-30856: Pre-0.3.0 versions are vulnerable to a tool-name collision and indirect prompt injection via an MCP client naming convention (mcp_{service}_{tool}), allowing a remote MCP server to hijack tool execution and potentially exfiltrate prompts/context or run other tools with use...

CVE-2026-30856

WeKnora is an LLM-powered framework designed for deep document understanding and semantic retrieval. Prior to version 0.3.0, a vulnerability involving tool name collision and indirect prompt injection allows a malicious remote MCP server to hijack tool execution. By exploiting an ambiguous naming...

CVE-2026-30856 WeKnora: Tool Execution Hijacking via Ambigous Naming Convention In MCP client and Indirect Prompt Injection

WeKnora is an LLM-powered framework designed for deep document understanding and semantic retrieval. Prior to version 0.3.0, a vulnerability involving tool name collision and indirect prompt injection allows a malicious remote MCP server to hijack tool execution. By exploiting an ambiguous naming...

CVE-2026-30856 WeKnora: Tool Execution Hijacking via Ambigous Naming Convention In MCP client and Indirect Prompt Injection

WeKnora is an LLM-powered framework designed for deep document understanding and semantic retrieval. Prior to version 0.3.0, a vulnerability involving tool name collision and indirect prompt injection allows a malicious remote MCP server to hijack tool execution. By exploiting an ambiguous naming...

WeKnora 安全漏洞

WeKnora is an open-source framework based on LLM developed by Tencent. It features deep document understanding using the RAG paradigm, semantic retrieval, and context-aware answers. Prior to version 0.3.0, WeKnora had security vulnerabilities. These vulnerabilities were caused by tool name...