SAFL is a research and capacity-building lab for advancing trustworthy AI in the world's middle powers.
Through our work, we are mapping the full landscape of work needed for countries like India to understand developments at the frontier of AI and develop the capacity to navigate ensuing challenges through research, institutions, and individual enterprise.
Study and mitigate the challenges from advanced AI and publish research findings.
Understand how AI diffuses across and reshapes society and the economy.
Incubate talent to tackle the world's most pressing problems.
Agentic AI is moving into India’s highest-stakes deployment surfaces faster than the evidence base needed to govern it. India needs original evaluations, red-teaming infrastructure, and attribution methods. That work is currently under-supplied for Indian languages, culture, and DPI.
The first extensive whitepaper covering red-teaming and evaluation of the latest Sarvam models. Building India-specific benchmarks to cover Sarvam's readiness and capability for Indian and global markets.
SAFL is building Secure AI Hubs across India - physical and hybrid co-working spaces where individuals interested in safe and trusted AI, AI security, and the societal and economic impacts of AI can synergize, synthesize, and their execute ideas.
We are looking for community builders and volunteers who would lead our hubs across India. If this resonates with you, we would love to hear from you.
Sign up Now →Since 2023, we have been building programs in safe and trusted AI. We ran India's first in-person AI safety careers fellowship, founded a global talent fellowship that places researchers into frontier safety labs, and published technical research on evaluating AI systems.
We built and ran the Global AI Safety Fellowship at SteadRise.
The fellowship surfaced technical and governance talent across 11+ global hubs including regions underrepresented in AI safety. Its recruitment funnel reached more than 5,200 candidates and produced 20+ offers, with 10 accepted placements. Through it, we contributed researchers to the Anthropic Fellowship, the UK AI Security Institute, and FAR.AI, etc.
We work with academia and policymakers to advance trustworthy AI, and contributed to the India AI Impact Summit 2026.
At the AI summit in New Delhi, we ran a workshop on hardware-rooted sovereignty featuring experts like Prof. Stuart Russell alongside government officials like India's IT Secretary. We also convened a pre-summit panel supported by the IndiaAI Mission on the sidelines of the AAAI-26 conference in Singapore.
We founded India's first alignment research and careers fellowships.
The first, an ML alignment fellowship at IIT Delhi in 2023, adapted BlueDot Impact's curriculum and incorporated capstone projects, career support, and mentorship. 22 people completed it, and within five months 20 of them had committed to a substantial alignment research project. The second, a residential alignment research fellowship hosted across Bengaluru and London with mentors from Mila-Quebec AI Institute and FAR.AI. The third, an AI safety careers course, which drew more than 600 applicants including top 10 JEE rankers, STEM researchers, and working professionals across South Asia.
We built talent infrastructure that maps more than 68,200 ML researchers and engineers globally.
We used it to place candidates into labs and grantee organizations doing frontier safety work, including UK AISI, FAR.AI, Apollo Research, GovAI, and Anthropic.
As SteadRise Alignment Research Fellows, we did technical research on evaluating LLM agents.
We had two technical AI safety research projects accepted across three NeurIPS workshops. We also contributed dangerous-capability evaluations to the UK AI Security Institute, including an implementation of Humanity's Last Exam added to its open Inspect Evals repository.
Co-convened at the AI Impact Summit 2026 with LucidComputing. Agenda on verifiable Safe & Trusted AI infrastructure - hardware-layer verification, India's DPDP Act, and compute supply chains.
Official pre-summit panel for the India AI Impact Summit 2026, held at SUTD during AAAI-26 week. Speakers from MDDI, NUS AI Institute, IIT Madras, UNSW, AIDXTech, and FAR.AI.
Global South hackathon in collaboration with Apart Research building for India-specific AI safety and evaluation problems.
A baithak-style convening near Bharat Mandapam alongside the AI Impact Summit — bringing together academics from IITD, IITM, IIITD, IIITH, UNSW, MBZUAI, and ASU.
We support exceptional and expert talent to find opportunities to contribute to AI Safety and Governance, Global Health and Wellbeing, and related impactful cause areas.
Open roles across research, engineering, and operations as we grow the lab.
Explore Careers →Short, focused engagements to tackle a specific high-impact problem alongside our team.
Special Projects Fellow →1:1 career conversations and resources from our team and experts working on Global Health and Wellbeing, at the intersection of transformative AI and social good, and AI governance at large.