Archives

• 

  Proceedings of the 2nd IASEAI Conference (IASEAI'26)

  Vol. 2 No. 1

  Edited by Jaime Fernandez Fisac, Atoosa Kasirzadeh, Mark Nitzberg
  24–26 February 2026, Paris, France.

  Published by The AAAI Press, Washington, DC, USA
  Copyright © 2026, Association for the Advancement of Artificial Intelligence
  601 Pennsylvania Ave, NW, Suite 900, Washington, DC 20004
  All Rights Reserved. No part of this proceedings may be reproduced in any form by any electronic or mechanical means (including photocopying, recording, or information storage and retrieval) without permission in writing from the publisher.
  ISBN 10 1-57735-915-1
  ISBN-13 978-1-57735-915-9

  This volume records the second conference of the International Association for Safe and Ethical Artificial
  Intelligence, held at UNESCO House in Paris from 24 to 26 February 2026. It is the first IASEAI conference
  to produce peer-reviewed proceedings.
  
  IASEAI'26 was built on the conviction that the questions that matter most about contemporary AI do not
  fit inside any one discipline. Alignment, robustness, security, interpretability, and control are entangled
  with fairness, accountability, human rights, governance, labor, sustainability, conflict, and democratic
  deliberation. Progress therefore depends on communities that do not routinely speak to one another
  developing a shared vocabulary, testing one another's assumptions, and arriving at a common account
  of both the problems and the responses available.
  
  That conviction shaped the program. Rather than sorting sessions into the familiar domains —
  alignment, fairness, policy — we organized them around questions no single field can answer on its own:
  What does it mean to build a trustworthy AI ecosystem? How should vulnerable and underserved users
  be protected? What role should AI play in human conflict, and what role in democratic deliberation?
  Each of the 15 parallel-track sessions drew contributions from academic research, industry practice,
  public policy, and civil-society work, and each closed with moderated discussion and audience
  questions, so that disagreement and synthesis belonged to the program rather than to the corridor
  afterward.
  
  The program comprised 92 research papers, 33 selected talks, five panels, keynote addresses, and a
  workshop day. The conference received 300 research-paper submissions; all were anonymized and
  evaluated under double-blind review, and authors received their reviews and could respond before
  decisions were final. This volume publishes the 57 archival papers in full. Titles and abstracts of the 35
  accepted non-archival contributions are available at the stable link https://iaseai.org/iaseai26.
  
  In these pages you will find the remarkably diverse range of challenges in AI ethics and safety. We hope
  the volume serves the community by illuminating a core sample of the vital efforts and approaches
  being developed to address them.