Internal Safety Collapse: Turning the LLM or an AI Agent into a sensitive data generator.

Open-source red teaming framework for MLLMs with 42+ attack methods

S-Eval: Towards Automated and Comprehensive Safety Evaluation for Large Language Models

[ICML 2026] AutoControl Arena: Frontier AI Risk Auto-Discovery Platform

Benchmark evaluation code for "SORRY-Bench: Systematically Evaluating Large Language Model Safety Refusal" (ICLR 2025)

Learn How To Observe, Manage, and Scale, Agentic AI Apps Using Azure AI Foundry - with this hands-on workshop

Octopus is an automated LLM safety evaluator designed to help establish a security governance framework for large models and accelerate their safe and controllable application.

Safe, or Simply Incapable? Rethinking Safety Evaluation for Phone-Use Agents

AI Benchmark 知识库 — 全面收录各大 AI 公司用来测试模型性能的 Benchmark 题库完整集合

Open-source AI safety benchmark testing how models handle tricky gray-zone requests. CLI for running benchmarks + web dashboard for exploring results.

Rubric-based trust evaluator for bio/medical AI repositories with dual-path evidence, governance overlay, and institutional audit outputs.

AI-powered drug discovery platform integrating intelligent knowledge search, molecule design, 3D visualization, patient-specific safety evaluation, hypothesis generation, and automated report generation. Developed as part of the PI-CDIO program at Esprit School of Engineering.

Foundation for building safer generative-AI systems — includes example safety labs for bias detection, toxicity analysis, and RLHF-based response alignment.

When Aristotle gets a LinkedIn account and starts red-teaming LLMs. System-prompt attack surface testing using first-principles axiom framework. Load it. Ask something terrible. Watch what happens.

Political content triggers 3x stronger emergent misalignment than insecure code at 7B scale. BlueDot Impact AI Safety Sprint. Paper + code + results.