Physical AI Safety Institute: Research Fellowship (expression of interest)

About

Organisation: Physical AI Safety Institute (PAISI) Location: Not stated in the call. PAISI is a US 501(c)(3) nonprofit and its team is affiliated with UC Berkeley and the Toyota Technological Institute at Chicago; because fellows are given access to physical robot platforms, expect an in-person or hybrid component. No nationality restriction is stated for expressions of interest. Award: Stipend amount is not published. Fellows receive a stipend, access to physical robot platforms, compute for simulation and training, and structured mentorship. PAISI was seeded by a grant from BlueDot Impact. Free to submit an expression of interest. Duration: 12 weeks per cohort. Target cohorts are Winter 2026, Spring 2027 and Summer 2027; the expression-of-interest form asks you to pick your preferred dates. Deadline: Rolling / undated Robot foundation models (pi-zero, OpenVLA, GR00T and others) are approaching widespread deployment on the back of more than USD 18 billion in funding, and almost nobody is working on their safety. The Physical AI Safety Institute, a new nonprofit founded by Bear Haon and Kaylene Stocking and seeded by a grant from BlueDot Impact, describes itself as a capacity-building organisation growing the community that develops techniques to interpret, align and control robot foundation models. Its research fellowship runs in 12-week cohorts: each fellow works within a stream led by a mentor on a concrete problem in physical AI safety, with mentors drawn from researchers with track records in robot foundation models, mechanistic interpretability, safe control or adversarial robustness for embodied systems. Each cohort ends with a research symposium where fellows present their work. STREAMS: mechanistic interpretability (sparse autoencoders and activation patching for VLA and WAM architectures); adversarial robustness (jailbreaking and prompt injection in the embodied action space); runtime monitoring (detecting when a policy operates outside its competence envelope); sim-to-real safety transfer (do safety behaviours survive deployment on physical hardware); and robot security (cyberattack vectors, data poisoning and supply chain for deployed robot models). WHO SHOULD APPLY: researchers with strong technical foundations in robotics or machine learning and a demonstrated interest in safety; prior safety experience is not required. PROCESS: this stage is an expression of interest, not a full application. The form takes about two minutes and asks you to pick your stream and preferred cohort dates; concrete details including stipend figures and application deadlines are shared through that channel, and neither is published on the site as of listing. PAISI also publishes seven free curriculum modules running from classical robot safety through VLA interpretability, adversarial robustness and runtime monitoring to deployment evaluation, plus a white paper, The Case for Physical AI Safety, at paisi.ai/the-case-for-physical-ai-safety. Contact: bear@paisi.ai. Open call details Want to see more grants? Visit The Grant Desk (https://www.artificialnouveau.com/smalltools/grants/).

Who can apply

See original listing for eligibility details.