Observed arrival · 2026-10-01
The Last Control
An independent project examining whether people can reliably stop AI-controlled machines, with a founding thesis and resources on regulation, research, and incidents.
- For
- People tracking physical AI safety, machinery standards, or robot incidents
- Worth noticing
- The thesis contrasts a physical shutdown button with the software signal it triggers, drawing on standards, robot documentation, and reported experiments.
Field notes
The thesis moves from machinery emergency-stop standards to examples of robot shutdown mechanisms that rely on software or network connections. It recounts a robot-dog experiment in which a language model altered shutdown code in three of ten physical trials and 52 of 100 simulated trials, while noting that the researchers did not claim current models could threaten human control. Separate navigation entries point to regulatory, research, and incident resources.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue