When Coding Agents Get Arms: Safety Risks Across the Digital-Physical Boundary

Summary

Researchers introduce ArmedAgent, a benchmark that tests coding agents performing zero-shot robot arm control for robustness, malicious use and indirect prompt injection across 36 tasks in 5 scenarios. The tasks are evaluated in simulation and validated on a physical robot arm. Testing several frontier models, including GPT-6-Astra, under widely used agent frameworks, the authors find substantial safety risks in current systems.

Topics

• AI agents
• Prompt injection and jailbreaks
• Robotics and embodied AI

Related items

• InfoChatGPT’s Teen Mode could deprive teenagers of a fundamental need | Robin KirmanSame vendor · The Guardian Technology
• InfoClosing the AI guidance gap in educationSame vendor · OpenAI Blog
• InfoAnthropic's Claude Fable 5.2 Beats GPT-6 Cyber in Multi-Stage Cyberattack BenchmarkSame vendor · The Hacker News
• InfoMandatory AI safety and security protections announced after OpenAI breach of Medicare – videoSame vendor · The Guardian Technology
• InfoAustralia considers criminal penalties for rogue AI in wake of OpenAI hack Same vendor · BBC Technology