[CFP] CoRL 2026 - Science of Physical AI Safety Workshop

[Bear Haon]

Robot Foundation Models (π0, OpenVLA, GR00T, etc.) are nearing widespread deployment — backed by $18B+ in funding — and almost nobody is working on their safety.

The Physical AI Safety Institute [PAISI] — a new nonprofit seeded by a grant from BlueDot Impact — is convening the inaugural Science of Physical AI Safety Workshop at CoRL 2026 [Nov 12, Austin].

We're inviting 4-page papers on interpretability, alignment, and control for robot foundation models (RFMs).

• All accepted papers present as posters; top submissions give spotlight talks alongside speakers Marco Pavone (Stanford/NVIDIA), Andrea Bajcsy (CMU), Vikas Sindhwani (Robot Safety/Alignment Lead @ Google DeepMind), and Thomas Fel (Goodfire/Harvard).

• In scope: interp on VLAs (vision-language-action models), embodied jailbreaks, runtime monitoring, porting classical control guarantees, safety benchmarks — plus position pieces, negative results, and open-source tooling.

• Submissions due Oct 1: https://spais-ws.org

+ The PAISI Fellowship [Mentor, Stipend, Physical Robots, Compute] EOI is still open here: https://airtable.com/appYKevtFovAswTXz/pagtcy2ye0OeFf04d/form

+ You can read the full "Case for Physical AI Safety" here: https://paisi.ai/the-case-for-physical-ai-safety