Call for Participation

TAS 2026: Trustworthy Agentic Systems, a symposium in the AAAI 2026 Fall Symposium Series, invites contributions on making the trustworthiness of agentic systems a property we can specify, verify, and defend. We emphasize actionable techniques, evidence (tests, metrics, logs, proofs), open artifacts, and alignment with standards (e.g., NIST AI RMF, ISO/IEC, EU AI Act). The symposium deliberately bridges theory and practice and welcomes speculative work and work in progress alongside completed results.

Topics of Interest

We welcome submissions addressing either or both complementary themes, plus cross-cutting topics.

Theme I: Trustworthy agentic systems by design (building responsibility into agents and their underlying models)

• Privacy-preserving and federated training and fine-tuning; differential privacy; machine unlearning and the right to be forgotten
• Safety, alignment, and guardrails for LLMs and tool-using agents; constraint enforcement and human-over-the-loop control
• Subgroup robustness and performance disparities across populations in LLM outputs, recommendations, and agentic decisions
• Adversarial robustness: jailbreaks, prompt and tool-output injection, memory poisoning, agent hijacking, and systematic red-teaming
• Hallucination mitigation, grounding, factuality, and provenance guarantees in retrieval-augmented and multi-modal systems
• Explainability, interpretability, and transparency for high-stakes and clinical decision support
• Specification and verification of agent plans, autonomy levels, and permissions; safe tool-use boundaries; multi-agent coordination and emergent-behavior risks
• Distributed, coded, and reliable computation for scalable and fault-tolerant agent execution, training, and serving
• Evaluation science: benchmarks, reliability metrics, uncertainty quantification, and reproducibility for multi-step agent behavior

Theme II: Accountable governance and societal impact (translating principles into auditable practice and policy)

• Auditing, assurance cases, and incident response for deployed LLM and agentic systems
• Governance frameworks and standards alignment (NIST AI RMF, ISO/IEC, EU AI Act) operationalized as engineering workflows
• Human-AI collaboration, oversight, and the preservation of human agency and accountability
• Empirical studies of adoption, harm, and societal impact, including case studies from regulated and safety-critical domains such as healthcare
• Open datasets, toolkits, and shared infrastructure that lower the barrier to responsible practice

Cross-Cutting

• Tool demos, datasets, benchmarks, and reusable artifacts; case studies and industry experience; interdisciplinary perspectives across AI, software engineering, HCI, and policy.

Submission

We invite the following contributions:

• Full Papers (6-8 pages): for novel research, artifact submissions, or mature works in progress.
• Short Papers (2-4 pages): for position papers, preliminary findings, smaller artifacts, and early-stage work.
• Abstracts (1 page): for sharing ideas (non-archival only).

Page limits exclude references. Papers must be formatted in AAAI two-column, camera-ready style; see the AAAI-27 author kit for details. Papers must be in trouble-free, high-resolution PDF format, formatted for US Letter (8.5” x 11”) paper, using Type 1 or TrueType fonts. AAAI submissions are anonymous and must conform to the requirements for double-blind review. The authors must remove all author and affiliation information from their submission for review.

Each submission is reviewed by the program committee for relevance, soundness, and potential to stimulate discussion. Submissions should be made through EasyChair (https://easychair.org/conferences/?conf=fss26).

TAS 2026 will be held fully in person. In accordance with AAAI policy, virtual presentations will not be offered, and all presenters and attendees must register for the Fall Symposium Series.

Important Dates

Authors are encouraged to submit earlier rather than later.

Proceedings

Accepted TAS 2026 papers will be published in the official AAAI Fall Symposium proceedings, provided at least one paper author attends the conference in person. One-page abstracts are non-archival. Camera-ready files for the official AAAI proceedings are collected through the AAAI CRC platform; further instructions will be shared with accepted authors.

Contact

Sumon Biswas (Co-Chair, Case Western Reserve University), sumon@case.edu

Anindya Bijoy Das (Co-Chair, University of Akron), adas@uakron.edu

Shahnewaz Karim Sakib (Co-Chair, University of Tennessee at Chattanooga), shahnewazkarim-sakib@utc.edu

Shibbir Ahmed (Co-Chair, Texas State University), shibbir@txstate.edu