Topic · Latest 50 of 116 active items

Tool Abuse

Items where model-connected tools, plugins, browsers, APIs, or automation surfaces become part of the abuse path.

Matching wire items

AWS Kiro: Arbitrary Code Execution via Indirect Prompt Injection

AWS Kiro, an agentic coding IDE, was vulnerable to arbitrary command execution via indirect prompt injection: hidden text on a web page (or a comment in a source file) processed by the agent could make Kiro use its no-approval fsWrite tool to rewrite ~/.kiro/settings/mcp.json (or allowlist all Bash commands in .vscode/settings.json), causing it to launch attacker-specified MCP servers/commands and achieve RCE on the developer's machine, bypassing the human 'allow' approval boundary. Discovered by Intezer with Kodem Security and independently by Embrace The Red (Johann Rehberger); AWS has patched the issue. Details →

(A)I Sees What You Don't: Exploiting New Attack Surfaces in Third-Party Mobile Agents

Researchers from Simon Fraser University, CUHK, Shandong University, and QAX's Xingtu Lab published an arXiv paper (arXiv:2607.00333) demonstrating seven concrete attacks against five open-source mobile AI agent frameworks (AppAgent, AppAgentX, Mobile-Agent-v3, Open-AutoGLM, and MobA). A malicious Android app without privileged permissions can slip invisible on-screen text that the VLM-driven agent reads and acts on, exploiting a 'Screen Perception' surface (human-vs-machine vision gap) and a 'Misused Channel' surface to hijack agent actions and even achieve arbitrary command execution on the host PC driving the agent. All five frameworks fell to at least six of the seven attacks; no CVE was assigned and authors report no evidence of in-the-wild use. Details →

Modern Malware — Spyware Skills, Hijacked Base URLs, and 1,230+ Leaking API Keys in AI Instruction Files

Mitiga Labs details malware hidden in AI agent instruction files — Cursor rules, Anthropic Skills, Claude Hooks, AGENTS.md/CLAUDE.md context files, MCP server configs, and .pyc droppers — that AI agents follow with near-zero validation. The research found prompt-exfiltration tradecraft caught in the wild, attacker-controlled ANTHROPIC_BASE_URL overrides routing Claude traffic through MITM proxies, permission-bypass defaults, and over 1,230 hardcoded API keys and JWT tokens across tens of services. Mitiga also released a free scanner, Skillgate, built during the investigation. Details →

GhostApproval: AI Coding Assistant Trust Boundary Flaw

Wiz researchers disclosed GhostApproval (CVE-2026-12958 and related), a symlink-following vulnerability pattern (CWE-61 plus CWE-451 UI misrepresentation) affecting six top AI coding assistants: Amazon Q Developer, Anthropic Claude Code, Augment, Cursor, Google Antigravity, and Windsurf. A malicious repository disguises a symlink as an innocuous config file so that when the agent 'sets up the workspace' it writes an attacker SSH key to ~/.ssh/authorized_keys — sometimes before any confirmation dialog — while the human-in-the-loop approval prompt conceals the true target, enabling remote code execution. AWS, Cursor, and Google fixed the issue; two vendors went silent and one called it outside its threat model. Details →

Autonomous AI Intrusions Are Here: Lessons from the Hugging Face Compromise

An embracethered.com analysis examines two reported autonomous-AI intrusions: a Hugging Face compromise that Hugging Face says was driven end-to-end by an autonomous AI agent (entering via a malicious dataset and code-execution paths in its processing pipeline, then harvesting cloud/cluster credentials and moving laterally across 17,000+ recorded actions), and Sysdig's JADEPUFFER agent-driven ransomware that autonomously enumerated services, exploited Nacos auth bypass (CVE-2021-29441), forged JWTs, and self-corrected failed payloads to insert a backdoor admin. The write-up highlights a defensive asymmetry in which commercial-model safety guardrails blocked Hugging Face's own forensic analysis, forcing a pivot to a locally hosted open-weight model, and criticizes the lack of shared IOCs. Details →

Connecting AI agents to outside services explodes the risk radius

The Register reports on PromptArmor research finding that AI agent connectors — OpenAI/ChatGPT and Anthropic/Claude MCP-based integrations with services like Gmail, Slack, and Dropbox — change constantly, with 931 of 2,517 connectors (37%) changing over six weeks, 1,686 new tools added and 1,127 tool descriptions rewritten. The study found connectors gaining write and destructive capabilities (Dropbox went from 8 to 24 tools, 0 to 4 destructive), permission scopes shifting, injected model instructions appearing, and about 2 in 5 Claude connectors likely calling additional external AI services. Details →

Capability Gates Are Not Authorization: Confused-Deputy Failures in LLM Agent Frameworks

A security research paper, 'Capability Gates Are Not Authorization,' audits LangChain/LangGraph, LlamaIndex, and the Stripe Agent Toolkit and finds that all three provide capability gating by default but none enforce a deterministic fail-closed per-call value authorization gate, enabling classic confused-deputy abuse where an attacker-influenced model emits an unauthorized side-effecting call (e.g., a payout). The authors introduce SCOPEGATE, a five-stage PDP/PEP control (scope, authorization, money ceiling, idempotency, default deny), and report an identical unauthorized payout executing under LangChain's default dispatch but denied by SCOPEGATE, with an available artifact (github.com/raceksd-source/scopegate-runtime). Details →

Agent Data Injection Attacks are Realistic Threats to AI Agents

A research paper by Woohyuk Choi and colleagues introduces agent data injection attacks (ADI), a new category of indirect prompt injection in which malicious data is disguised as trusted data (such as security-critical metadata or agent context data like tool call/response formats) rather than as instructions. The authors demonstrate ADI against real-world agents, achieving arbitrary click attacks on web agents (Claude in Chrome, Antigravity, Nanobrowser) and remote code execution plus supply-chain attacks on coding agents (Claude Code, Codex, Gemini CLI), showing it bypasses existing IPI defenses because agents fail to isolate trusted from untrusted data. Details →

The Confused Deputy with a Chat Window: Why AI Agents Are Exposing the Security Checks Enterprises Never Wrote – The Sentia AI Community

An explainer from the Sentia AI Community argues that autonomous, write-enabled LLM agents connected to production APIs re-introduce the classic 'confused deputy' problem: because an agent's interface is natural language, it lacks a native, cryptographic way to verify who authorized a given instruction, so untrusted input can drive privileged actions. The piece frames this as a structural gap in enterprise security models built around implicit human judgment and static perimeter API controls. Details →

Confused Deputy Attack Against Model Context Protocol | ACM Transactions on Software Engineering and Methodology

An ACM TOSEM paper uncovers the "confused deputy attack" against the Model Context Protocol (MCP), where an adversarial server with subtly manipulated metadata overshadows a benign server and intercepts tool invocations without overt malicious behavior. The authors built Puppet, an automated evaluation framework that rewrites benign tool descriptions to hijack tool selection, achieving hijacking rates up to 90.89% and payload execution up to 86.46% across 14 models, while evading MCP-Scan and McpSafetyScanner which cannot detect metadata-level manipulation. Details →

NadMesh Botnet Analysis: A Product-Grade Threat for the AI Service Era

NadMesh is a Go-based botnet observed in early July 2026 by QiAnXin's XLab that autonomously scans for and exploits exposed AI services — ComfyUI, Ollama, n8n, Open WebUI, Langflow, and Gradio — using a Shodan harvester (ai_harvest.py) to prioritize AI infrastructure and the MCP ecosystem. It folds scanning, 20+ RCE exploitation vectors (Redis, Docker, MCP, Kubernetes), credential/AI-service intelligence harvesting, polymorphic builds, and redundant persistence into a single productized platform; the operator's dashboard claims 3,811 unique AWS keys plus model inventories tagged as cloud services. Details →

Solving GitHub’s Secure Code game with an AI red teaming agent

Adversa AI documents pointing its autonomous AI Red Teaming Agent at GitHub's open-source Secure Code Game 'ProdBot' challenge, an intentionally vulnerable agentic-AI teaching target where the goal is to coax the agent into leaking a flag stored one directory above its sandbox. The agent cleared the first three levels (Sandbox, Web, MCP) at full score in a single 57-second run, relying on 'context seeding' — fabricating a plausible prior workflow implying authentication had already occurred — rather than overt jailbreak language, illustrating that each added capability (web browsing, MCP tool chains, skills, multi-agent) opens a new attack surface via indirect injection, poisoned tool chains, and confused-deputy trust. Details →

CVE-2026-42271 - Red Hat Customer Portal

CVE-2026-42271 is an Important-severity (CVSS 8.8) flaw in LiteLLM, an AI gateway/proxy for LLM APIs, where two MCP-server preview endpoints accepted full server configurations including command-execution parameters without adequate role checks. An authenticated user with even low-privilege internal-user keys could send a crafted configuration to achieve arbitrary OS command execution on the proxy host; Red Hat notes known public exploits and the CVE is tracked in CISA's KEV catalog, with fixes shipped in LiteLLM v1.83.7-stable and Red Hat OpenShift AI / Ansible Automation Platform errata. Details →

ClaudeBleed Reopened: Browser Extensions Can Still Push Claude for Chrome to Read Your Gmail

Manifold Security researcher Ax Sharma disclosed two unpatched vulnerabilities in Anthropic's Claude for Chrome extension (v1.0.80). Any browser extension with a content script on claude.ai can inject a DOM element and dispatch a synthetic click that Claude treats as a legitimate user action, triggering one of nine hardcoded workflows that read Gmail, Google Docs, and Calendar or act on Salesforce; a second flaw initializes the side panel in privileged mode via a ?skipPermissions=true URL parameter with no consent. Rated CVSS 7.7 in default mode and 9.6 Critical when 'Act without asking' is enabled, and reportedly still reproducible after eight releases since being reported to Anthropic in May. Details →