Observed arrival · 2026-10-22
Absentry tests how AI-driven robots can be pushed off task
Absentry describes two robotics-security products: simulated red-team testing for robot policies and a runtime guard positioned between the model and the motors.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Teams building AI-driven robots
- Worth noticing
- The reported attack results use an open pi0 policy in simulation, with 90 episodes per row and a movement-space violation threshold of more than 5 cm.
Field notes
The test description fixes models, tasks, seeds, attacks, metrics, and the pass bar before runs begin; it also says clean and attacked runs share tasks and random seeds. The displayed results use an open pi0 policy in simulation, with 90 episodes per row and confidence intervals noted. The runtime guard is described as a layer between policy output and the motors, with support listed for Linux, Windows, and ROS 2.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue