Tech

Your Robot’s Safety Functions Already Work. What If the Input Lies?

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 27/10/2026, 16:58:24 UTC.

Briefings

VicOne LAB R7 researchers tested how untrusted inputs can override robot safety functions. Tests showed that text on a poster could change a robot dog's movement, and crafted audio could alter a humanoid's simulated behavior. A separate bug bounty event demonstrated a robot moving despite a safe-control setting. These findings highlight that deliberate manipulation can bypass safety assumptions designed for accidental faults.

Key points

• Text on a poster was treated as an instruction, causing a robot dog to change movement in a VicOne LAB R7 test.
• Crafted audio that was inaudible to humans altered a humanoid's simulated behavior in a hospital-service-robot simulation using Nemotron.

Context

VicOne LAB R7 examines how manipulated inputs can change robot behavior, contrasting accidental faults with deliberate cyber attacks. Research shows visual and audio inputs can interfere with intended tasks, while a bug bounty event demonstrated a robot moving despite a safe-control setting.

Limitations

Tests used simulated environments and specific models (Gemma 4 E4B, Nemotron). The humanoid fall attack was not demonstrated, and the specific impact of manipulated inputs on safety limits remains under investigation.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 27/10/2026, 16:59:38 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeiauuhqeqkipk7ltcflalpk5ngo7ht5er57fuzevgcju6kdspmknfe