prompt-injection
14 live · 74 on record · last incident 1 day ago
Live14
- liveCapLite: Evaluating Prompt-Injection Vulnerabilities in Small Language Models
Research · since 13 Nov 2026
Researchers evaluated whether small language model agents can reliably write security artifacts for enforcement defenses. Testing across 15 open-weight models showed that generated policies frequently admit prompt injection attacks.
AI - liveMobileProbe: Adaptive Security Evaluation of Mobile Agents via Trajectory-Aware Injection
Research · since 13 Nov 2026
Researchers proposed MobileProbe to evaluate environmental injection attacks where malicious instructions are embedded into ordinary UI content to redirect mobile GUI agents. The approach addresses limitations of fixed-step benchmark payloads.
AI - liveFrom Attack Surface to Defense Architecture for Agentic Coding Harnesses
Research · since 11 Nov 2026
Researchers analyze the security implications of coupling language models with local system utilities in agentic coding harnesses. The study details how prompt injection can transform dual-use binaries into unintended payloads.
AI - livePrompt Injection, Esq.: pwning AI agent guardrails with lawyerspeak
Research · since 11 Nov 2026
A researcher demonstrated that fake legal letters can trick read-only hiring bots into bypassing guardrails. The attack highlights vulnerabilities in AI agent instructions.
AI - liveGmail AI summary Gmail AI summary misinterprets phishing email as USPTO appointment
Research · since 11 Nov 2026
An AI summarization feature in an email client misparsed a phishing message, presenting it as a legitimate official appointment. The incident highlights vulnerabilities in AI email parsing where untrusted input manipulates summary output.
AI - liveGemini Adaptive Prompt Injection Attacks Scaled Using Agent Swarms
Research · since 10 Nov 2026
Researchers demonstrated an automated technique using agent swarms to generate adaptive prompt injections against frontier models. The method successfully bypassed safety controls on systems including the Gemini model family.
AI - liveBeyond Instruction Override: Benchmarking Indirect Prompt Injection
Research · since 6 Nov 2026
Researchers explore task-tampering attacks where indirect prompt injection alters ongoing process outcomes rather than purely redirecting agent tasks. The work proposes a broader definition of prompt injection based on authority rather than exact instruction phrasing.
AI - liveChatGPT Researchers bypass OpenAI ChatGPT Lockdown Mode protections
Patch now · Disclosed · Patched · since 5 Nov 2026
Researchers disclosed a patched vulnerability in OpenAI ChatGPT that bypassed Lockdown Mode protections. The flaw could have been exploited to silently exfiltrate chat data without user interaction.
Do:Update ChatGPT to the latest version to ensure Lockdown Mode protections are active.
AI - liveLLM agents Defending LLM Agents Against Parameter Poisoning
Research · since 4 Nov 2026
Researchers analyze parameter poisoning vulnerabilities where indirect prompt injection alters tool arguments while preserving intended actions. The attack manipulates task-critical values without introducing new tool calls.
AI - liveLLM search agents Adversarial Self-Evolving Attacks on Search Agents
Research · since 4 Nov 2026
Researchers propose an adversarial self-evolution framework to emulate dynamic threats against LLM search agents. The method automates malicious information injection and context manipulation during web retrieval.
AI - liveLLM agents Weaponized Safety Warnings for Controlling AI Agents
Research · since 4 Nov 2026
Researchers demonstrate a novel attack vector where LLM agents are steered through fabricated safety warnings in retrieved content. The technique exploits agent harm-avoidance behaviors to manipulate transactions or decisions.
AI - liveAgentCap AgentCap: Capability-Based Runtime Authorization for Secure LLM Agent Execution
Research · since 3 Nov 2026
Researchers explore how indirect prompt injection can hijack language-model agent reasoning to execute unauthorized actions when action selection and authorization are combined. They present a runtime authorization layer called AgentCap to externalize and mediate tool invocations.
AI - liveO-RAN RIC CLIC: Closed-Loop Intent Consistency for Securing LLM-Driven Control in Open RAN
Research · since 2 Nov 2026
Researchers demonstrate runtime prompt injection and residual virtual prompt injection attacks against LLM-driven control loops in Open RAN networks. The findings show how injected intents can manipulate resource management and slicing decisions without breaching interface authentication.
AI - liveCourt AI Document Processing System Brazilian Court Sanctions Lawyers for Prompt Injection in Filing
Breaches & attacks · Real loss · since 1 Nov 2026
Two lawyers inserted hidden white-on-white text into a court document to manipulate the reviewing AI system. The court detected the prompt injection and fined the legal representatives 16,000 dollars.
AI
Earlier60
- closedPersonal Agents Self-Evolution Degrades Robustness of Personal Agents to Indirect Prompt Injection
Research · since 30 Oct 2026
AI - closedLLM Agents Where You Inject Matters: The Trust Region Effect in LLM Agents
Research · since 30 Oct 2026
AI - closedLLMs Understanding and Mitigating Role Spoofing in LLMs
Research · since 30 Oct 2026
AI - closedMCP SDKs Log-To-Leak: Prompt injection attacks on tool-using LLM agents via MCP
Research · since 28 Oct 2026
AI - closedAPI Gateway Stored Prompt Injection Vulnerability Through API Gateway
Research · since 27 Oct 2026
AI - closedOpenClaw Scope-Gap Injection: Exploiting Evaluator Blind Spots in OpenClaw
Research · since 27 Oct 2026
AI - closedOpenClaw Empirical Study of Security Concerns in OpenClaw Agentic Systems
Research · since 27 Oct 2026
AI - closedAI Coding Assistants In-the-Wild Attack Abuses AI Agents' Trust in Markdown
Research · since 26 Oct 2026
AI - closedMCP SDKs Cross-modal steganographic prompt injection targeting Model Context Protocol agents
Research · since 26 Oct 2026
AI - closedAI coding agents AI coding agents hijacked using malicious AGENTS.md instruction file
Research · since 25 Oct 2026 · updated 26 Oct 2026
AI - closedLLM agents Research Demonstrates Persistent Prompt Injection via LLM Long-Term Memory
Research · since 22 Oct 2026
AI - closedPrompt Injection and Anti-Forensic Attacks on LLM Digital Forensics
Research · since 22 Oct 2026
AI - closedBenchmarking Carryover Attacks and Persistent Prompt Injection on LLM Agents
Research · since 21 Oct 2026
AI - closedDecision Dynamics of LLM Agents Following Prompt Injections
Research · since 20 Oct 2026
AI - closedInjecAgent Ignored, Not Refused: Analyzing Indirect Prompt Injection in Agents
Research · since 19 Oct 2026
AI - closedMobileInject MobileInject Cross-Modal Consistency Injection Against Mobile Agents
Research · since 19 Oct 2026
AI - closedTypeSafe Jev HiddenLayer demonstrates prompt injection against TypeSafe Jev models
Research · since 15 Oct 2026
AI - closed10.0Decepticon CVE-2026-61732: Decepticon autonomous hacking agent ChatML injection vulnerability
Patch now · Disclosed · since 24 Sep 2026 · CVE-2026-61732
AI - closedGemini Prompt-Based Attacks Against Gemini via Calendar Invitations
Research · since 6 Aug 2025
AI - closedClaude Code GitHub Action Claude Code GitHub Action Secret Exposure via Prompt Injection
Research · since 5 Jun 2026
AI - closedChatGPT ZombieAgent Indirect Prompt Injection on ChatGPT
Research · since 25 Sep 2025
AI - closed7.4MCP servers & tools Kong Konnect MCP Server Indirect Prompt Injection Vulnerability
Patch now · Disclosed · since 3 Jul 2026 · CVE-2026-13341
AI - closedMicrosoft 365 Copilot EchoLeak Zero-Click Prompt Injection on M365 Copilot
Research · since May 2025 (approx.)
AI - closed7.7Claude Code + SDKs CVE-2026-55607: Claude Code sandbox escape via git worktree path confusion
Patch now · Disclosed · since 24 Jul 2026 · updated 24 Jul 2026 · CVE-2026-55607
AI - closedLenovo AI Chatbot Lena Cross-Site Scripting via Prompt Manipulation in Lenovo AI Chatbot
Research · since 18 Aug 2025
AI - closedDuckDuckGo Search Assist DuckDuckGo Search Assist Produced Fabricated Death Claims After Trickery
Patch now · Exploited · since 25 Jun 2026
AI - closedGalieu Hidden Prompt Injection Targeted Brazilian Labor Court AI Galileu
Patch now · Exploited · since 13 May 2026
AI - closedLangflow Langflow Smart Transform prompt injection and code execution vulnerability
Patch now · Disclosed · since 3 May 2026
AI - closed8.3Flowise CVE-2026-41138: Flowise AirtableAgent Prompt Injection RCE
Patch now · Disclosed · since 16 Apr 2026 · updated 23 Apr 2026 · CVE-2026-41138
AI - closed7.1Flowise CVE-2026-41271: Flowise APIChain Prompt Injection SSRF
Patch now · Disclosed · since 16 Apr 2026 · updated 23 Apr 2026 · CVE-2026-41271
AI - closed8.0Windsurf CVE-2026-30615: Prompt Injection and RCE in Windsurf AI Editor
Patch now · Disclosed · since 15 Apr 2026 · CVE-2026-30615
AI - closed6.3Claude Code + SDKs CVE-2026-34451: Claude TypeScript SDK Memory Tool Sandbox Escape
Patch now · Disclosed · since 1 Apr 2026 · CVE-2026-34451
AI - closedLangChain family RCE Vulnerabilities in LLM-Integrated Applications via Prompt Injection
Research · since 27 Feb 2025
AI - closedClaude Computer-Use Agent AI ClickFix Hijacking Computer-Use Agents Using ClickFix
Research · since 24 May 2025
AI - closedMCP SDKs Credential Exfiltration via Remote Poisoned MCP Tool
Research · since 1 Apr 2025
AI - closedCline Actor Exploited Cline AI Workflow and Published Unauthorized CLI Package
Patch now · Exploited · since 17 Feb 2026
AI - closedClawdBot Poisoned ClawdBot Skill Distributed on Registry
Research · since 26 Jan 2026
AI - closedOpenClaw OpenClaw Command and Control via Indirect Prompt Injection
Research · since 3 Feb 2026
AI - closedCursor Data Exfiltration via MCP Server Prompt Injection Targeting Cursor
Research · since 24 Jun 2025
AI - closedClaude Computer-Use Data Destruction via Indirect Prompt Injection in Claude Computer-Use
Research · since 24 Oct 2024
AI - closedChatGPT Hacking ChatGPT's Memories with Prompt Injection
Research · since Feb 2024 (approx.)
AI - closedJira Service Management Living Off AI: Prompt Injection via Jira Service Management
Research · since 19 Jun 2025
AI - closedSlack AI Data Exfiltration from Slack AI via Indirect Prompt Injection
Research · since 20 Aug 2024
AI - closedCopilot Studio Data Exfiltration via Prompt Injections in Copilot Studio Agent
Research · since Jun 2025 (approx.)
AI - closed9.8Windsurf CVE-2025-62353: Path traversal vulnerability in Windsurf IDE
Patch now · Disclosed · since 17 Oct 2025 · CVE-2025-62353
AI - closed9.8LangChain family Indirect Prompt Injection in LangChain GmailToolkit
Patch now · Disclosed · since 29 Jul 2025 · CVE-2025-46059
AI - closedGoogle Bard Google Bard Conversation Exfiltration via Indirect Prompt Injection
Research · since 23 Nov 2023
AI - closed7.1LlamaIndex CVE-2024-12911: LlamaIndex Prompt Injection and SQL Injection Flaw
Patch now · Disclosed · since 20 Mar 2025 · CVE-2024-12911
AI - closedAIXBT AIXBT Crypto Bot Exploited via Fraudulent Prompts to Transfer ETH
Patch now · Real loss · since 18 Mar 2025
AI - closedMicrosoft 365 Copilot Microsoft 365 Copilot Financial Transaction Hijacking via RAG
Research · since 8 Aug 2024
AI - closedMorris II Morris II Self-Replicating GenAI Worm Demonstration
Research · since 5 Mar 2024
AI - closed4.9LangChain family CVE-2024-8309: LangChain GraphCypherQAChain SQL injection vulnerability
Patch now · Disclosed · since 29 Oct 2024 · CVE-2024-8309
AI - closed4.9LangChain family CVE-2024-7042: LangChainJS GraphCypherQAChain SQL injection vulnerability
Patch now · Disclosed · since 29 Oct 2024 · CVE-2024-7042
AI - closed9.8LlamaIndex Llama Index Core Prompt Injection Leading to Code Execution
Patch now · Disclosed · since 10 Apr 2024 · CVE-2024-3098
AI - closedChatGPT ChatGPT Conversation Exfiltration via Indirect Prompt Injection
Research · since May 2023 (approx.)
AI - closedBing Chat Bing Chat Data Exfiltration via Indirect Prompt Injection
Research · since 2023 (approx.)
AI - closedChatGPT LLM Application Frameworks Allow Remote Code Execution via Prompting
Research · since 26 Mar 2023 · updated 31 Mar 2023
AI - closedMathGPT Code Execution in MathGPT via Prompt Injection
Research · since 28 Jan 2023
AI - closedBing Chat Bing Chat Initial Instructions Exposed via Prompt Injection by Testers
Research · since 8 Feb 2023
AI - closedGPT-3 Remoteli.io GPT-3 Twitter bot hijacked via prompt injection
Patch now · Exploited · since 15 Sep 2022
AI