Call for Papers
Topics
We encourage paper submissions relevant to (but not limited to) the following topics:
Survey workflow-level defenses (e.g., sandboxing, provenance tracking) to prevent data leakage and unauthorized actions
Explore model evaluation protocols and compliance criteria inspired by the EU AI Act’s conformity assessments (risk assessments, technical documentation, logging) and investigate the role of independent third-party audits and “safety labels” for AI systems.
Analyze liability frameworks for harm caused by misaligned or compromised AI agents and evaluate the implications of cross-border coordination, drawing on cooperation insights on technical AI safety collaboration among geopolitical rivals
Present the “AI Scientist” paradigm, which advocates for non-agentic, uncertainty-aware models to avoid runaway autonomy and discuss design patterns (e.g., provable separation of planning and execution, human-in-the-loop gates) that enforce safe operating envelopes.
Important Dates
All deadlines are due 23:59 PM in GMT time zone.
- Submission Deadline: Aug 29, 2025
- Acceptance Notification: Sep 22, 2025
Submission Instruction
This workshop is non-archival. The review process is double-blinded. Submissions should be anonymized appropriately.
Abstracts and papers can be submitted through OpenReview.
Format
We welcome both short papers no longer than 4 pages and long papers of up to 9 pages, excluding references and unlimited supplementary materials. Please use the RegML @ NeurIPS 2025 template and submit your paper(s) in PDF format.