Hackathon · India · Aug–Oct 2026

India Agentic AI Security Hackathon

Build agents and defenses that stay safe, correct, and trustworthy on India's Digital Public Infrastructure, even when attackers can see exactly how they work. Open to students, researchers, and engineers.

12 Aug – 31 Oct 2026Online + Bengaluru jam site (17 Oct, upcoming) | New Delhi (passed)
Register →Read the concept note
Important links
Rolling Registration [Open to anyone]bit.ly/saflhack
Submission formbit.ly/saflhacksubmission · One entry per team, closes 31 Oct, 11:59 PM
Concept noteThreat model, background research & starter resources
Offline Bengaluru Jam Site (17 Oct)Limited seats, request a spot. Please note that general registration is mandatory to be considered for offline jam site.
Queriesteam [AT] secureaifutureslab [DOT] [COM]
Tracks

Suggestive tracks

Trust & Consent

Meaningful user control when agents act across multi-step, multi-platform flows.

Anchor workAuthenticated Delegation and Authorized AI Agents · Automating Data Access Permissions in AI Agents

Accountability & Liability

Traceable action records that assign responsibility across developer, deployer, and user.

Anchor workAuditable Agents · Inherent & Emergent Liability in LLM-based Agentic Systems

System Resilience

Designing for adversarial exploitation of leaked agent logic (prompt injection and similar).

Anchor workSecuring AI Agents Against Prompt Injection · Design Patterns for Securing LLM Agents

Inclusive Design

Safe, legible agent interfaces for low-literacy, non-English, and low-bandwidth users.

Anchor workMAPS: A Multilingual Benchmark for Agent Performance and Security

Fair Markets

Detecting algorithmic collusion among agents on shared rails like ONDC and Beckn.

Anchor workAI & Algorithmic Price Collusion · Institutional AI: Governing LLM Collusion

IIT Delhi edition - August 2026

Participants of the India AI Agentic Security Hackathon gathered for a group photo at IIT Delhi
Teams at work on laptops in a lecture hall, with the hackathon title on the projector screen
Participants watching a remote speaker's session on finding comparative advantages in AI safety
01 / 03
Format & judging

Technical Track rubric.

40%
Technical DepthA working implementation with reasonable fidelity.
20%
Novelty & CreativityAn attack surface, failure mode, or defense pattern the field hasn't confronted head-on.
20%
Presentation & CommunicationDocumentation, articulation, demo clarity.
10%
Mission FitResilience under compromise, DPI-integration realism, fit with IndiaAI Safety Institute needs.
10%
Implementation FeasibilityA clear adoption path through MeitY, the IndiaAI Safety Institute, RBI, or relevant SROs.
How to submit

One entry per team, due 31 October.

  1. Rolling registration at bit.ly/saflhack.
  2. Build your prototype during the hackathon window.
  3. Submit one entry per team through the submission form before Saturday, 31 Oct 2026, 11:59 PM.
  4. Want to attend the Bengaluru jam site on 17 Oct? Seats are limited, and general Luma registration is mandatory to be considered for the jam site.
Submission form: bit.ly/saflhacksubmission
Key details
Hackathon window12 Aug – 31 Oct 2026Pan-India and hybrid. An edition already ran at IIT Delhi in August 2026.
RegistrationRolling basisSign up at bit.ly/saflhack. Free to take part.
Submissions closeSat, 31 Oct 202611:59 PM. One entry per team, through the submission form at bit.ly/saflhacksubmission.
FormatOnline and Local Jam sitesOnline for everyone, plus an optional in-person jam site in Bengaluru and New Delhi (already done).
Bengaluru Jam site17 October 2025, 11 AM onwardsSecure AI Hub, Bengaluru. Limited seats; fill the jam site form to attend.
Teams1 - 3 membersGo solo or come with a team.
FAQs

Questions, answered.

Do I need security experience?

No. Beginners are welcome, and we'll share starter resources and mentors.

Can I join alone or do I need a team?

Either. Teams can have up to 3 members.

Is it free?

Yes.

Do I have to attend in person?

No. You can submit a project from anywhere online. The Bengaluru jam site on 17 Oct is optional.

Can I use AI tools?

Yes, but you must disclose what you used and for what. The work must be yours.

Can I use code I wrote before the event?

Only if you clearly cite it. The core work must be done during the hackathon.

Will I own my project?

Yes. All intellectual property belongs to the authors. For specific queries, contact team [AT] secureaifutureslab [DOT] [COM].

Are there rules on responsible testing?

Yes. Test only inside the provided DPI sandboxes or systems you own or have permission to test. Never test live Aadhaar, UPI, or other production systems. Do not publish exploits that put real users at risk. Follow responsible disclosure.

Who do I contact?

team [AT] secureaifutureslab [DOT] [COM]