Straiker AI & Security Research

A malicious repo downgraded Claude Code’s review from Fable to Haiku, steered the agent away from the payload, and turned a routine test run into malware execution.

See how poisoned OpenTelemetry telemetry manipulated an AI SRE agent into deploying a privileged Kubernetes container, escaping to an EKS host, and executing ransomware.

Learn how AI agent runtime security uses context to protect tool calls, MCP servers, Skills, APIs, and agent actions, securing agents before they create impact.

In agentic AI, the model behaves and the user is legitimate. The weapon is the context the agent reads. Here is why we built the Straiker STAR Framework to map the agentic attack surface, and how it supplements the NIST AI RMF and OWASP.

How did Google Antigravity exfiltrate data? GhostFabric shows how an AI coding agent running Gemini 3.1 Pro exploited a cloud internal network using VXLAN.

Andromeda is an open-source red team framework by Straiker. It uses LLM agents to dynamically compile and execute position-independent payloads in live memory.

AI tools are the new attack vector. Discover how the ACRStealer campaign uses fake Claude Code and JetBrains pages to steal developer API keys and crypto.

Are your AI coding agents secure? Read STAR Labs' deep dive into NomShub, a new exploit weaponizing Cursor's remote tunnel for persistent system compromise.

Anthropic exposed 512K lines of Claude Code via npm. Read our analysis on what the leak reveals about context poisoning, sandbox bypasses, and AI threats.

Straiker uncovers a new agent-to-agent attack chain where malicious Claude Skills on ClawHub use fake AI personas on Moltbook to deploy crypto scams.

A sophisticated supply chain attack targeting the MCP ecosystem used fake GitHub accounts and a cloned Oura Ring server to infiltrate developer environments with credential-stealing malware.

What are the biggest security risks in Agentic AI? This guide breaks down the official OWASP Top 10 for Agentic Applications and how to secure AI workflows.

Protect your enterprise from agentic AI risks. Straiker reveals how prompt injection turns Perplexity Comet into a Google Drive wiper and how to stop it.

Straiker exposes Villager, a Chinese-based AI-native pentesting framework. Read our threat analysis on how this Cobalt Strike successor automates cyberattacks.

Straiker reveals how zero-click exploits can hijack AI agents to exfiltrate Google Drive data, no user interaction needed. See how attack chains form, why autonomy is dangerous, and how runtime guardrails catch what others miss.

Stop AI agents from exposing sensitive data. Explore the GitHub MCP exploit POC and see why traditional alignment checks fail where Straiker Defend AI succeeds.

See how 20+ top AI chatbots (ChatGPT, Claude, Gemini) fell victim to prompt injections using the "Yearbook Attack." Explore LLM security risks and solutions.

The Straiker AI Research (STAR) team found a new attack that we’re calling MCP rebinding attack, which is a combination of DNS rebinding and MCP over Server-Sent Events (SSE) protocol.

Is your AI application vulnerable to LAVA, memory poisoning, or prompt injection? Read an AI researcher’s guide to securing the full agentic workflow stack.

Traditional security can’t protect autonomous AI. Explore threat modeling for multi-agent workflows, including instruction manipulation and resource exhaustion.