Safe Human FutureSubmit Abstract
Back to conference

SAFE HUMAN FUTURE FOUNDATION

Call for
Presentations

Data Privacy Governance • AI Governance

The Safe Human Future Foundation invites abstract submissions of up to 1,000 words, preferably with an accompanying demo, for presentations at Safe Human Future 2026 at Hilton Mumbai International Airport Hotel on 16 October 2026.

THE MANDATE

Research, practice, and regulation into shared results.

The event advances the Foundation’s mandate to convert research, field practice, and regulatory insight into shared, publishable results.

Work is organised through two standing Working Groups — Data Privacy Governance and AI Governance — and their Technology Research, Industry, and Policy Subgroups.

Submissions are organised around four topic families. The focus topics are indicative of where the field is heading, not exhaustive; contributions from research, industry, and policy perspectives are equally welcome.

PREFERENCE FOR ABSTRACT + DEMO

Bring the research to life.

The organising committee strongly encourages submissions to include a demonstration alongside the written abstract. Where submissions are of comparable quality, preference will be given to abstract + demo submissions over abstract-only submissions.

Demonstrations may show working systems, prototypes, tools, or reproducible experiments for a cross-disciplinary audience.

PRESENTATION

Selected authors will present at Safe Human Future Mumbai 2026.

Presentations will reach researchers, industry practitioners, and policy specialists from the Foundation’s Technology Research, Industry, and Policy Subgroups.

Registration fees are waived for accepted authors.

Travel allowance and accommodation may be provided to students unable to obtain funding.

THE RESEARCH FRONTIER

Four families. One safer human future.

01

Safe Quantification

Privacy Unit · Working Group I — Data Privacy Governance

Towards a common, defensible “unit” of privacy that can be measured, composed, and audited across systems and jurisdictions.

  1. Privacy threat modeling and quantification — structured frameworks for identifying, categorising, and scoring privacy threats across the data lifecycle.
  2. Standardization of privacy metrics — common definitions, benchmarks, and reporting formats across tools, vendors, and jurisdictions.
  3. Quantifying leakage in frontier generative models — measuring memorization, extraction, and membership-inference risk with reproducible audits.
  4. Verifiable, regulatory-grade privacy metrology — attestable budgets, third-party-checkable accounting, and standardized privacy scoring.
02

Safe Data

Zero-Knowledge Proofs · Working Group I — Data Privacy Governance

Proving facts about data and computation without revealing the data itself — the succinct-proof frontier now shaping IACR venues.

  1. Zero-knowledge verifiable credentials — selective disclosure, unlinkable presentation, and revocation without exposing identity data.
  2. Post-quantum and transparent succinct proofs — lattice- and hash-based SNARKs/STARKs, folding and accumulation schemes, and lookup arguments.
  3. Zero-knowledge machine learning — proofs of correct inference and training, plus dataset provenance, licensing, and compliance.
  4. Zero-knowledge proofs for data rights — client-side proving, proof-of-personhood, and privacy-preserving consent, eligibility, and KYC/AML proofs.
03

Safe AI

Mechanistic Interpretability · Working Group II — AI Governance

Explaining the internal computation of frontier models and turning that understanding into safety evidence.

  1. Scalable feature decomposition — sparse autoencoders, transcoders, and cross-layer attribution graphs for human-auditable features.
  2. Causal and weight-based interpretability — circuit discovery through interventions, activation patching, parameter decomposition, and training dynamics.
  3. Keynodes and training — tracking influential circuits and features as they emerge, stabilise, or shift during training.
  4. Interpretability for safety assurance — detecting deception, hidden objectives, and sandbagging through regulator-grade evaluations and audits.
04

Safe Agents

Agentic Privacy · Bridging Working Groups I & II

Preserving privacy and accountability as autonomous agents plan, remember, delegate, and act over sensitive data and systems.

  1. Red teaming for agentic privacy leaks — adversarial evaluation of unintended disclosure through tool use, memory, and multi-step planning.
  2. Privacy-preserving agent memory — architectures enforcing minimal disclosure, purpose limitation, and forgetting.
  3. Cryptographic agent identity and scoped delegation — verifiable credentials, authenticated protocols, and accountable delegation.
  4. Agentic contextual integrity and proofs — formal frameworks and cryptographic proofs that data flows respect purpose and consent.

ABSTRACT SUBMISSION INSTRUCTIONS

Submit through the form.

Abstracts of up to 1,000 words should be submitted via the Safe Human Future submission form, preferably accompanied by a live or recorded demo. Authors can submit and revise their abstracts through the form until the deadline, with a maximum of two abstracts per author, as a single author or with co-authors.

Dual Submissions & Prior Work: Because the conference is non-archival, we welcome submissions that are under review at other conferences or journals, as well as work that has recently been published or presented elsewhere.

Submission form: Open the submission form

WHAT TO INCLUDE

Abstract · Demo link or materials · Author and co-author detailsPresentation materials, including slides and accompanying demo materials, should be submitted through the form ahead of the event.

PRESENTATION & PUBLICATION

Bring your work to a cross-disciplinary audience.

Authors of selected abstracts will be invited to present at Safe Human Future Mumbai 26, sharing their work with researchers, industry practitioners, and policy specialists drawn from the Foundation’s Technology Research, Industry, and Policy Subgroups.

ABSTRACT INSTRUCTIONS

1,000 words maximum

Abstracts of up to 1,000 words are submitted through email. Presentation materials, such as slides for in-person talks, should be submitted through the same form ahead of the event.

PROCEEDINGS OPTIONS

Full-Paper OptionUp to 4 pages, including tables and figures, excluding references. The template, detailed instructions, and copyright terms will follow upon acceptance. Original, unpublished work only.
Abstract OptionYour original submitted abstract may be included as open-access front matter in the Safe Human Future Mumbai 26 Proceedings.

SELECTED ABSTRACTS & NON-ARCHIVAL PROCEEDINGS

Retain full ownership of your research.

Safe Human Future Mumbai 2026 is a non-archival venue. We welcome work that is currently under review, planned for future submission, or recently published elsewhere.

OPTION 01

Extended Abstract / Working Paper (Non-Archival)

Authors may submit an extended version, up to 4 pages, using the Safe Human Future template for distribution to conference attendees. Submissions will be distributed under a non-exclusive license, allowing you to submit or publish your full paper at other venues without restrictions. Authors retain full copyright of all submitted abstracts and presentation materials.

OPTION 02

Abstract-Only Entry

Authors may choose to have only their submitted 1,000-word abstract included in the non-archival conference program.

Presentation slides and demo materials should also be submitted via email ahead of the event.

SUBMISSION DEADLINES

Mark the dates.

015 Oct 2026Abstract submission via the form
027 Oct 2026Selection of papers for presentation
0310 Oct 2026Author confirmation
0416 Oct 2026In-person presentation
0515 Dec 2026Submission of extended paper (optional)
0610 Jan 2027Notification to authors (extended paper)
0728 Jan 2027Paper Publication

QUESTIONS?

Speak with the Programme Committee.

For questions or further information, contact the Programme Committee.

Email the committee