Understand this story
On November 20, 2026, Ant Group's inclusionAI open-sourced LingSante MedGuard, a medical AI guardrail model designed to review medical AI answers for safety, compliance, and ethical risks. It supports rule comprehension, numerical reasoning, evidence verification, error localization, and multimodal hallucination detection. The accompanying LingSante MedGuardBench contains 14,588 risk-detection records and 578 images; the 27B model achieved a Full F1 score of 87.4% and a Text F1 score of 86.6% on the original experimental suite. The project uses supervised fine-tuning and multi-teacher on-policy knowledge distillation, while its code, original annotations, and adapted content are licensed under Apache-2.0. There are currently no other evidence-backed developments.
AI-generated from public reports
Latest development · Report
Ant Group’s inclusionAI Open-Sources LingSante MedGuard, a Medical AI Guardrail ModelDevelopments
1 confirmed developments
- Source:English text is being prepared.
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 20
Ant Group inclusionAI: New GitHub RepositoryFirst-party
Ant Group’s inclusionAI Open-Sources LingSante MedGuard, a Medical AI Guardrail Model
Ant Group’s inclusionAI has open-sourced LingSante MedGuard, which reviews medical AI responses for safety, compliance, and ethical risks and supports rule comprehension, numerical reasoning, evidence verification, error localization, and multimodal hallucination detection. LingSante MedGuardBench contains 14,588 risk-detection records and 578 images; the 27B model achieved 87.4% Full F1 and 86.6% Text F1 on the original evaluation suite. The project uses supervised fine-tuning and multi-teacher on-policy knowledge distillation. Its code, original annotations, and adapted content are licensed under Apache-2.0.
AI Trends