Aim and Scope

The goal is to build reliable and secure machine learning (ML) models that are resilient in adversarial and evolving settings, including those introduced by emerging Artificial General Intelligence (AGI).

As ML systems advance toward greater autonomy and generalization, new challenges in security, robustness, and privacy become increasingly critical. AGI amplifies concerns around distributional shifts, adversarial inputs, long-term reliability, and value alignment. These raises pressing questions: How can we design models that are robust to novel and adversarial inputs, especially in open-world scenarios? How can systems detect, reason about, and adapt to distributional shifts or unforeseen operational contexts? When can we trust intelligent systems to act reliably across extended time horizons and new domains? How do we embed safety, transparency, and ethical behaviour in systems that learn and evolve beyond their initial training? Recent progress in adversarial ML, privacy preservation, and robust optimization offers a strong foundation, but must evolve to meet these challenges.

We aim to bring together researchers from diverse fields such as reinforcement learning, AGI safety, human-robot interaction, game theory, cognitive science, and cybersecurity to advance reliable and trustworthy learning. Our focus areas include robustness (e.g., adversarial attacks, data poisoning), trustworthiness (e.g., explainability, transparency, privacy), and scalability (e.g., generalization to novel tasks and objectives).

This task force promotes recent theoretical and empirical advances in secure learning, drawing on past insights to meet the growing demands of intelligent and adaptive systems.

Task Force Chair:

• Owen Vallis, OpenAI, USA (owensvallis@gmail.com)

Task Force Vice-Chairs:

• Leo Zhang, Griffith University, Australia (leo.zhang@griffith.edu.au)

• Sherin Mathews, US Bank, USA (s.mathews217@gmail.com)

Task Force Members:

• Catherine Huang, Google, USA (catherinehuanglei@gmail.com)

• Wenjian Luo, Harbin Institute of Technology, Shenzhen, China (luowenjian@hit.edu.cn)

• Huiyu (Joe) Zhou, University of Leicester (hz143@leicester.ac.uk )

• Tayo Obafemi-Ajayi, Missouri State University, USA (tayoobafemiajayi@missouristate.edu)

• Taehong Kim, Chungbuk National University (taehongkim@cbnu.ac.kr)

• Viet Vo, Swinburne University of Technology (vvo@swin.edu.au)

• Jinyu Tian, Macau University of Science and Technology (jytian@must.edu.mo)

• Yanjun Zhang, University of Technology Sydney (Yanjun.Zhang@uts.edu.au)

• Manish Bhatt, OWASP (manish.bhatt13212@gmail.com)

• Shuren Qi, The Chinese University of Hong Kong (shurenqi@cuhk.edu.hk)

• Wanlun Ma, Swinburne University of Technology (wma@swin.edu.au)

• Celeste Fralick, McAfee LLC, USA (celeste_fralick@mcafee.com)

• Yufei Chen, City University of Hong Kong (yufeichen8-c@my.cityu.edu.hk)

• Chao Chen, RMIT University (chao.chen@rmit.edu.au)

• Di Wu, La Trobe University (D.Wu@latrobe.edu.au)