AI Safety Mixer: Embodied AI | SAFL
Robots are moving out of the lab and into factories, warehouses, roads, hospitals and homes. More and more, they run on the same kind of AI models that power chatbots. When a chatbot gets something wrong, you get a bad answer. When a robot gets something wrong, it acts in the real world, next to real people.
This Saturday, Secure AI Futures Lab is hosting an AI Safety Mixer on embodied AI at our Secure AI Hub in Bengaluru. We’re bringing together people working across academia, research labs, and startups to discuss robotics and secure AI.
Over two hours, we’ll circle around one theme: what it takes for robots and other embodied AI systems to be safe for the people around them. The evening will flow like this:
Welcome and the theme → Meet and Greet → Open discussion → One-on-one networking over light refreshments
Come to exchange ideas and meet like-minded people working on similar problems.
📅 Date: Saturday, 24 October
🕔 Time: 6:00 PM IST onwards
We’ll share the exact address once your registration is approved.
---
Suggested reading before you come
Nothing here is required, but each piece adds context for the evening:
Start here: It’s Surprisingly Easy to Jailbreak LLM-Driven Robots (IEEE Spectrum, 2024). A short, readable account of researchers tricking AI-controlled robots, including a robot dog, into ignoring their safety rules.
Technical: Jailbreaking LLM-Controlled Robots (arXiv, 2024). The paper behind that story, which introduces RoboPAIR, an automated attack that succeeded against all three robot systems it was tested on.
Governance: Embodied AI: Emerging Risks and Opportunities for Policy Action (Perlo, Robey, Barez, Floridi and Mökander). A map of the physical, informational, economic and social risks from embodied AI, the gaps in current US, EU and UK rules, and what testing, certification and liability rules could look like.
Right after: AI Futures Dialogue, 7 PM, online
SAFL’s speaker series continues with Dr. Roshni Lulla, Co-Founder and Head of Research at the Institute for Humane Robotics. IHR is an independent research and standards body working on how robots coexist with people, and Roshni’s research spans affective neuroscience, moral cognition and misalignment. You’re welcome to join from the hub, or register online at bit.ly/afdreg.
Hosted by Secure AI Futures Lab, a research and capacity-building lab for advancing trustworthy AI in the world’s middle powers.