Responsible Federated LLMs via Safety Filtering and Constitutional AI

Eunchung Noh; Jeonghun Baek

Responsible Federated LLMs via Safety Filtering and Constitutional AI

Computation and Language 2026-05-19 v2 Distributed, Parallel, and Cluster Computing Multiagent Systems

Authors: Eunchung Noh , Jeonghun Baek

Abstract

Recent research has increasingly focused on training large language models (LLMs) using federated learning, known as FedLLM. However, responsible AI (RAI), which aims to ensure safe and trustworthy responses, remains underexplored in this context. In FedLLM, client-side training data may contain harmful content, resulting in unsafe LLMs that can generate inappropriate responses. Aggregating such models into a global model and redistributing it to clients risks the widespread deployment of unsafe LLMs. To address this, we incorporate two well-established RAI techniques into FedLLM: safety filtering and constitutional AI. Our experiments show that these methods significantly improve LLM safety, achieving over 20% improvement on AdvBench.

Keywords

federated learning large language models safety filtering constitutional AI AdvBench

Cite

@article{arxiv.2502.16691,
  title  = {Responsible Federated LLMs via Safety Filtering and Constitutional AI},
  author = {Eunchung Noh and Jeonghun Baek},
  journal= {arXiv preprint arXiv:2502.16691},
  year   = {2026}
}

Comments

Accepted at the 6th Workshop on Trustworthy NLP (TrustNLP), ACL 2026

Responsible Federated LLMs via Safety Filtering and Constitutional AI

Abstract

Keywords

Cite

Comments

Related papers