InfoNewsLLM-specific
AI-Defined Vehicles: Securing the Decisions That Drive Vehicle Behavior
- Published
- Record updated
Summary
Researchers in the ReasonBreak study, published in May 2026, used corrupted textual inputs to manipulate trajectories generated by NVIDIA's Alpamayo driving models, increasing collision rates in closed-loop simulation. The source also describes a black-box adversarial patch study on vision-language driving models and a reflection-based backdoor study against DriveLM that increased inference latency. The authors frame these as research results rather than demonstrated exploits against production vehicles.
Related items
- HighCVE-2026-75612: yt-mcp arbitrary file read through add_attachment file_pathSimilar attack · NVD/CVE Database
- HighCVE-2026-130121: The AI Puffer - MCP Server for Claude & ChatGPT, with AI Chatbot plugin for WordPress is vulnerable to Stored…Similar attack · NVD/CVE Database
- InfoExternal sandboxes for coding agentsSimilar attack · Simon Willison's Weblog
- MediumResearchers hijack Hugging Face account for popular Kokoro AI modelSimilar attack · BleepingComputer
- InfoEvolution Under Fire: Evolving Agent Safety Harnesses with Adaptive Injection StressorsSimilar attack · Arxiv (cs.CR + cs.CL + cs.LG)