28 Zero-Day Findings in Under Two Days. Is Robotics Security Keeping Pace?
spectrum.ieee.org Nov 16, 2026

28 Zero-Day Findings in Under Two Days. Is Robotics Security Keeping Pace?

AI-summarised brief · reviewed before publication

VicOne disclosed 28 zero‑day vulnerabilities discovered in under two days through AI‑assisted analysis of software that underpins the Physical AI ecosystem, including tools from Google, NVIDIA and Meta. All findings are currently classified as zero‑days, though some lack CVE identifiers pending vendor review. The flaws span the robotics lifecycle—development and training, simulation and validation, and deployment/runtime—with 75 % affecting pre‑deployment stages. Exploits could corrupt pretrained models, falsify simulation results, or disrupt remote‑stop commands, potentially altering robot behavior or obscuring operator control. VicOne’s research highlights how AI can accelerate vulnerability discovery, giving attackers a similar advantage and leaving robotics teams with limited time to patch or mitigate risks.

💡 Why It Matters

  • · Accelerated AI‑driven discovery compresses the window between vulnerability emergence and remediation, forcing robot developers to confront security threats before physical deployment.