AAAI Fall Symposium: Engineering Safety-Critical AI Systems 2025
November 6-8 | Arlington, Virginia
Engineering Safety-Critical AI Systems
Artificial intelligence (AI) has increasing application to high-risk settings, but foundational practices for engineering safe AI systems remain few, and research in safety engineering for AI remains scattered across disparate fields of study. The AI community needs more discussion on identifying, assessing, and mitigating AI hazards. In this symposium, we will strike at a single fundamental question: How should we build AI systems for safety-critical applications? The symposium will bring together three communities as part of a broader effort to advance the discipline of engineering AI for safety: (1) domain experts who want to use AI in safety-critical applications, (2) AI researchers, engineers, and practitioners who build AI capabilities, and (3) safety and systems engineers who are tasked with designing, developing, and testing systems that include AI components
Topics
While most topics around safety and AI are welcome, we are especially interested in topics that inform how to engineer AI systems now and in the future. Areas of interest include but are not limited to the following:
- Safety requirements engineering for AI and/or new safety standards for AI systems
- Software architectures for increased AI system safety
- Uncertainty quantification and/or robustness in AI components or systems
- Methods for defining AI system specifications
- Safety test and evaluation of AI components or systems
- Software tooling to support safety engineering in AI
- Reporting of high impact failure modes or cases in AI systems
- High-risk application domains of AI that require specific definitions of safety
- Case studies of safety engineering for deployed AI systems
- Formal specification and verification of AI systems and/or processes for certifying AI system safety
- Provably safe AI and safe-by-design AI
- Risk assessment of generative AI in safety-critical systems such as medical robots and autonomous vehicles
- Human-AI interaction in safety-critical systems
Symposium Agenda
Day 1, November 6
9:15 Welcome and Opening Remarks: Symposium Organizing Committee
9:30 Keynote: Thomas Dietterich, Oregon State University
10:30 Break
11:00 Spotlight Talks: Authors of Selected Papers
12:30 Lunch
2:00 Keynote: Laura Freeman, Virginia Tech National Security Institute
3:00 Spotlight Talk: Authors of Selected Papers
3:30 Break
4:00 Panel: Moving AI Research to AI Safety Engineering (Moderator: Eric Heim) | Participants: Thomas Dietterich, Oregon State University; Laura Freeman, Virginia Tech National Security Institute, Zoe Szajnfarber, George Washington University)
Day 2, November 7
9:15 Day 2 Opening Remarks and Schedule: Symposium Organizing Committee
9:30 Keynote: Benjamin Grosof, DARPA
10:30 Break
11:00 Posters, Authors of Selected Papers
12:30 Lunch
2:00 Keynote: Patrick Hall, George Washington University
3:00 Break
4:00 Panel: Identifying & Mitigating AI Hazards (Moderator: Mary Cummings)
Day 3, November 8
9:15 Day 3 Opening Remarks and Schedule: Symposium Organizing Committee
9:30 Panel: Industry Perspectives on AI Safety (Moderator: Lu Feng) | Participants: Mauricio Castillo-Effen, Lockheed Martin; Josephine Lamp, Dexcom)
10:30 Break
11:00 National Study on AI Engineering Outbrief: Matthew Gaston, Carnegie Mellon University Software Engineering Institute (CMU SEI)
Call for Participation
Format
The symposium will be two and a half days, featuring invited keynote speakers, selected paper presentations, and panels.
Submission Requirements
There are two tracks for submission:
Track #1: Research and Development – This track seeks papers articulating new AI safety research or technical reports describing new safety engineering artifacts (process, procedure, standards, software architectures, tooling, etc.). Preference will be afforded to works that are rigorous, have evidence to support real-world application, and advance the discipline of engineering safe AI systems.
Track #2: Case Studies in Engineering AI Systems – This track seeks papers that highlight the practical engineering challenges with building safe AI systems in a challenging application domain and/or present a case study in engineering safe AI systems in the real world. Preference will be afforded to works that showcase real-world problems with high impact and rigorous and well-justified engineering solutions to these problems.
For both tracks, we are seeking papers in two formats: (a) short papers (2-4 pages, excluding references) and (b) full papers (6-8 pages, excluding references). All submissions must be formatted using the AAAI-25 author kit and be submitted through the AAAI EasyChair site. All papers will undergo single-blind review (the papers do not need to be anonymized). Accepted full papers will have options to be included in the AAAI symposium proceedings.
- Paper submission deadline: August 11
- Notification of acceptance or rejection: August 22
- Camera-ready deadline: August 29
Symposium Committee
Co-chairs
Wanyi Chen (Duke University, wc151@duke.edu)
Dr. Eric Heim (Carnegie Mellon University, etheim@sei.cmu.edu)
Organizing Committee
Dr. Gregory Canal (Johns Hopkins University, Greg.Canal@jhuapl.edu)
Dr. Mary Cummings (George Mason University, cummings@gmu.edu)
Andrew Dolgert (Carnegie Mellon University, ajdolgert@sei.cmu.edu)
Dr. Lu Feng (University of Virginia, lf9u@virginia.edu)
Dr. Ritwik Gupta (University of Maryland and University of California, Berkeley, ritwikgupta@berkeley.edu)
Chase Midler (CrowdStrike, cmidler@gmail.com)
Dr. Sanjeev Mohindra (MIT Lincoln Laboratory, smohindra@ll.mit.edu)
Oren Wright (Carnegie Mellon University, owright@sei.cmu.edu)
The AAAI Fall Symposium series is usually held on the east coast in late October/early November. It offers an intimate setting where participants can share ideas and learn from each other’s artificial intelligence research. Topics for the symposia change each year, and the limited seating capacity and relaxed atmosphere allow for workshop-like interaction.
Learn more about the 2025 Fall Symposium Series sponsored by the Association for the Advancement of Artificial Intelligence