Our spaces across India

Secure AI Hubs

Physical and hybrid spaces where people working on safe and trusted AI, AI security, and AI impacts can synergise, synthesise, and execute their ideas. Our hubs are operating across New Delhi and Bengaluru with offices in Hyderabad and Chennai coming soon.

Saket, New DelhiKoramangala, BengaluruGachibowli, Hyderabad (coming soon)TIDEL Park, Chennai (coming soon)
Hub 01

New Delhi

Delhi is the capital of India and home to key administrative and decision-making bodies. SAFL's Delhi Hub is the apex center inviting researchers, academicians, government officials, and policymakers to collaborate and discuss ideas in AI security and governance. Our hub serves as the physical embodiment of the community we have nurtured for years. Between weekly reading groups, hackathons, mixers, and much more, the space is also selectively available for individuals or organizations for work.

Saket, New Delhi

Participants discussing a paper at the SAFL Delhi hub in Saket
Technical AI Governance Reading Group in session at the SAFL Delhi hub
The reading group room at the SAFL Delhi hub in Saket
Technical AI Governance Reading Group at the Delhi hub, Saket.
Hub 02

Bengaluru

Bengaluru puts the hub beside the country's densest concentration of AI engineers and startups. Co-located with leading AI startups, our cozy Koramangala space runs as co-working with hybrid programming - talks, workshops, and webinars that brings together the people researching about fundamental AI security problems.

Koramangala, Bengaluru

The SAFL Bengaluru hub in Koramangala
Programs

Running Programs

  • Technical AI Governance Reading Group (Aug 2026 - Oct 2026)New Delhi · In person
    Sign up

    August 2026 – October 2026. Weekly on Fridays and Saturdays.

    We hold two batches at our Delhi Hub, which can be attended independently: Fridays, 4–6 PM and Saturdays, 11 AM – 1 PM.

    Overview

    You can't really govern what you can't measure. Aviation and cars didn't get safe from rules alone, but when we built ways to actually check them: crash tests, black boxes, incident reports. Powerful AI systems don't have much of that yet. Technical AI governance is the field that is building this.

    A growing body of work now describes where governance falls short, but a much smaller set of proposals talks about the tools that might address it. Each week we take one such paper and examine how well its argument holds up.

    Structure of each Session

    In every session:

    • We begin with 30 minutes of reading.
    • We whiteboard the paper's main claims and the evidence behind each.
    • We open the discussion onto broader implications, scrutinizing the proposal through a middle power, India-focused lens.
    • We work through the strongest objection we can find and the strongest defence against it, and close by drawing out the key takeaways and where to read next.

    What We Cover

    Each week we take one paper and examine how well its argument holds up.

    Prerequisites

    No advanced preparation is needed. There is nothing to read in advance, and no prior background is assumed. We introduce the paper so newcomers and returning participants begin on equal footing.

    The group brings together researchers, policy experts, engineers, and builders to unpack each idea, interrogate our assumptions, and understand the conditions under which it succeeds or fails.

    Selection Policy

    Invite only. Once you have filled out the form and explained your fit, we will send you an email confirmation regarding your attendance and the exact location of the group. Please note that we reserve the right to refuse entry if you miss more than two sessions.

    Cost

    This program is free of cost. Due to high demand for our programs, we request that you try to attend all sessions as we generally have a waiting list.

    Questions?

    Reach out at team@secureaifutureslab.com

    Sign up
  • India AI Agentic Security Hackathon (Aug 2026 - Oct 2026)New Delhi · Bengaluru · Hyderabad · Chennai · Hybrid

    Rolling hackathon over three months starting from August.

    Overview

    India's DPI—UPI, Aadhaar, DigiLocker, ONDC—was built for synchronous, human-confirmed interactions. AI agents now plan across time, call external services, hold memory, and act with limited oversight. They break that assumption at every step. When agents transact and delegate at population scale on infrastructure that never anticipated them, the failure modes are new, and the tools to address them do not yet exist in the Indian context. This hackathon starts building those tools.

    How it works

    1. Register and form a team of one to three.
    2. Work over several weeks — long enough for serious work rather than a rushed prototype.
    3. Attend offline Jam Sites in Delhi, Bangalore, Hyderabad, Chennai and elsewhere for mentorship and collaborators as they are announced.
    4. Submit online from anywhere in India, or virtually from anywhere in the world.

    What we cover

    Submission Tracks:

    • Technical — a working prototype or technical research addressing a real DPI security problem.

    Every challenge is grounded in a real DPI guardrail. Solutions are judged not on whether they work, but on whether they hold up when the agent's logic, prompts, and weights are public knowledge. We assume compromise and reward resilience.

    Selection Policy

    Open to anyone with a serious interest in AI safety and governance. Register via the form below. Jam Site attendance is optional but strongly recommended.

    Cost

    This hackathon is free to register and attend.

    About SAFL

    Secure AI Futures Lab is South Asia's first research and capacity-building lab for navigating powerful AI, working across AI governance and impacts research, expert talent advancement, government engagement, and ecosystem support. Foundational work supported by Schmidt Sciences and the Future of Life Institute, among others.

    Get involved

    Register: bit.ly/saflhack

    Institutions wanting to amplify this among students, organizations interested in associating, and practitioners who would like to contribute as mentors, judges, or speakers — write to us.

    Questions? team@secureaifutureslab.com

Next

Upcoming Hubs

SAFL is planning to open hubs across major hubs across Hyderabad, Chennai, and other major cities in India. Please join our Whatsapp channels, express your interest via the form below, and engage with our community to keep yourself updated on new developments.

Existing and upcoming hubs are open on all four counts — run your program, host a fellowship, propose an event, or take subsidized co-working. Tell us which one you want and we'll take it from there.

SAFL Hub Programs

Events and Programs

Programming runs across technical institutions in New Delhi, Bengaluru, Hyderabad, Chennai, and others - combining in-person hackathons with webinars, reading groups, mixers, seminars, expert talks, and hands-on workshops delivered both in person and virtually at each location. You can also run your program out of a hub, host a fellowship, propose an event, or use the space as subsidized co-working.

Hackathons & Technical Sprints

Live, in-person research sprints. Participants gather at the hub on a fixed jam day to build agentic AI systems together under a shared governance and safety lens — how these systems are architected, where they fail, and what oversight they need.

Co-working spaces

Subsidized desk space at the Delhi and Bengaluru hubs for people and teams working on safe and trusted AI - drop in for a day or take a standing seat, in the same room where the reading groups, workshops, and talks run.

AI Futures Dialogue | Expert Talk Series

Practitioners and researchers in direct engagement with students at each location - building understanding of trustworthy and agentic AI well beyond the people who show up for a hackathon.

Hands-on workshops

Working sessions on evaluations, red-teaming, and agent oversight, delivered in person and virtually at each location.

Reading Groups

Reading groups on topics related to safe and trusted AI, AI security, AI impact and more - a place to synergise, synthesise, and execute ideas.

Webinars & Seminars

Held both in person at the hub and virtually, so a session reaches a national audience rather than only the people who could make it into the room that day.

Join us at a hub.

Come to work, come to read, come to build. Run your program, host a fellowship, propose an event, or take subsidized co-working. We are also looking for community builders and volunteers to lead hubs across India — the same form reaches us either way.

Sign up for Hubs