India has one of the largest pools of AI engineers in the world. Very few have a clear path into safety work.
Why talent is the constraint
India’s public digital infrastructure (UPI, Aadhaar, DBT, ONDC) runs some of the largest multi-agent systems anywhere, with almost no safety research examining them. We think talent is the binding constraint. We train people, then help them find where they can do the most good, whether that is with us or somewhere else.
How we work
Based in India, open globally.
Approach
Simple
Clear, structured entry points into hard AI safety ideas.
Execution
Practical
Cohorts, working groups and research projects.
Outcome
Useful
Alumni have gone on to SPAR, MLCommons, AI Safety Camp, BlueDot Impact and Sentient Futures.
2Reading and mechanistic interpretability groups50 participants
3Technical AI Safety and AI Governance cohorts60 to 80 participants
4Research working groups5 groups
5Research programmeStarting 2027
The numbers, 2025 against 2026
2025 actuals against 2026 targets
Stage
2025
2026 target
Reached
~300
1,000
Trained in cohorts
30+
60 to 80
In AI safety research or job roles
15+
30 to 40
Research working groups
1
5
Active local groups
3
6
1:1 career advising sessions
~50
100
How we count
A person counts as in AI safety research or job roles if they hold a role, fellowship or facilitator position at a safety organisation, or have published safety work. Taking a course, reading, or intending to switch does not count.
What would make us stop
Fewer than 5 of 50 people from a cohort do anything in AI safety within 12 months.
Cohorts drift into generic AI ethics because we cannot fill them with people who care.
Values
Serve the journey, not the organisation
We are one stop on someone’s path into AI safety, not the destination. When an alum outgrows us and moves on, that is success, not attrition.
Say the uncomfortable truth
We give and receive direct feedback. The mission matters more than comfort or politeness.
Build in the open
We publish our outcomes, our reasoning and our mistakes, so others can check our work and build on it.
Collaborate, don’t compete
India’s AI safety field is small. We would rather grow it with other groups than compete for a fixed share.
From a reading group to two tracks and two cities.
Our history
2022
EA NIT Agartala is founded
It runs its first reading groups on AI safety.
2025
AISIN is founded in April
Cohort 1 trains 30+ participants. Within 12 months, about half have moved into AI safety roles or further fellowships.
2026
Two tracks, two cities, a second cohort
Separate Technical AI Safety and AI Governance tracks launch. City meetups run in Bengaluru and Delhi. Cohort 2 opens, and EA IIT Madras starts.
The team
Founder and Programme Director
Aditya Raj
AI safety researcher working on multi-agent safety and interpretability. He started AISIN in April 2025.
Background
SPAR Research Fellow, researching capability spillovers from safety techniques
Cooperative AI researcher and NeurIPS reviewer
Facilitator on BlueDot Impact’s Technical AI Safety course
