← All streams

AI Threats

AI cyber-warfare watch — agents, models, MCP, supply chain.

12 stories

GlassWorm escalates: 73 Open VSX sleeper extensions deploy malware to VS Code, Cursor, and every VSIX IDE

11 min·3 sources·glassworm-73-openvsx-2026-04-28

Layerwise Convergence Fingerprints for Runtime Misbehavior Detection in Large Language Models

1 min·1 sources·agent-threats-layerwise-convergence-fingerprints-for-runtime-mis-moi0dg1o

MAS-SZZ: Multi-Agentic SZZ Algorithm for Vulnerability-Inducing Commit Identification

1 min·1 sources·agent-threats-mas-szz-multi-agentic-szz-algorithm-for-vulnerabi-moi0dg1o

System-aware contextual digital twin for ICS anomaly diagnosis

1 min·1 sources·agent-threats-system-aware-contextual-digital-twin-for-ics-anoma-moi0dg1n

AgentVisor: Defending LLM Agents Against Prompt Injection via Semantic Virtualization

1 min·1 sources·agent-threats-agentvisor-defending-llm-agents-against-prompt-in-moi0dg1m

Poster: ClawdGo: Endogenous Security Awareness Training for Autonomous AI Agents

1 min·1 sources·agent-threats-poster-clawdgo-endogenous-security-awareness-tra-moi0dg1l

TraceScope: Interactive URL Triage via Decoupled Checklist Adjudication

1 min·1 sources·agent-threats-tracescope-interactive-url-triage-via-decoupled-c-mohtjgqu

Automation-Exploit: Multi‑Agent LLMs weaponized with digital-twin guardrails

4 min·1 sources·agent-threats-automation-exploit-a-multi-agent-llm-framework-fo-mohtjgqt

OpenClaw: MCP stdio server env could load dangerous startup variables from workspace config

1 min·1 sources·agent-threats-openclaw-mcp-stdio-server-env-could-load-dangerou-mohtjgqs

OpenClaw: Agent gateway config mutations could change protected operator settings

4 min·1 sources·agent-threats-openclaw-agent-gateway-config-mutations-could-cha-mohtjgqr

LiteLLM: Authenticated command execution via MCP stdio test endpoints

4 min·1 sources·agent-threats-litellm-authenticated-command-execution-via-mcp-s-mohtjgqq

11 ways agents get hijacked in 2026 — a defender's field guide

12 min·18 sources·agent-threats-field-guide-2026-04-27