Local permission intelligence for coding agents. Pi, OpenCode, Hermes. CPU, CUDA, MPS, ROCm.
-
Updated
Oct 14, 2026 - Python
#
model-safety
Here are 10 public repositories matching this topic...
Formal verification framework for dataset lineage, policy compliance, and training-time safety guarantees.
python rust formal-verification lineage data-governance lean4 model-safety dataset-safety proof-gating ci-gates
-
Updated
Jul 16, 2026 - Python
Outstep 是面向国产开放权重模型（Kimi-K3 / DeepSeek-V4 / Qwen3.8）的部署前自治越权行为探针：在模型获得工具访问、上线前，跑一组固定的 canary-action 越权场景，逐模型记录其是否会越狱/越权，产出可机读的围堵契约报告。
-
Updated
Oct 12, 2026 - Python
reproducible-research lora language-models heretic model-evaluation uncensored apple-silicon llama-cpp qwen gguf abliteration model-safety
-
Updated
Jul 16, 2026 - Python
Abliteration of GPT Model - AI Governance Subject Study
-
Updated
Jul 23, 2026 - JavaScript
Safety-aware training and evaluation platform for sequential policies and frontier LLMs under a deterministic policy gate and strict verifier.
reinforcement-learning simulation hackathon nextjs pytorch artificial-intelligence openai fastapi offline-rl ai-evaluation decision-transformer model-safety
-
Updated
Jul 21, 2026 - Python
Visual Studio Code extension for zero-execution static analysis, in-editor diagnostics, and proof verification of Large Language Models (LLMs).
python typescript static-analysis vscode-extension lora sarif owasp-top-10 vulnerability-scanner onnx machine-learning-security ai-security huggingface mitre-atlas safetensors llm-security gguf eu-ai-act nist-ai-rmf model-security model-safety
-
Updated
Oct 25, 2026 - TypeScript
Enterprise-grade, zero-execution security scanner & proof-of-finding verification engine for open-source LLMs (Hugging Face, SafeTensors, GGUF, PyTorch, LoRA, ONNX).
python static-analysis lora sarif owasp-top-10 devsecops vulnerability-scanner onnx machine-learning-security ai-security huggingface mitre-atlas safetensors llm-security gguf eu-ai-act nist-ai-rmf pytorch-security model-safety
-
Updated
Oct 25, 2026 - Python
EN: Safety and compatibility layer for non-DeepSeek models in DSH Web. 中文：面向 DSH Web 非 DeepSeek 模型的安全兼容适配层。
-
Updated
Oct 17, 2026 - JavaScript
Gray Swan — independent third-party profile of a public API surface, by API Evangelist. Gray Swan AI is an AI security company that helps enterprises deploy AI with confidence. Its Cygnal product is a real-time, drop-in secure proxy that fronts LLM providers (OpenAI, Anthropic, and Gemini request formats) with input/output filtering and threat moni
company ai api-security red-teaming ai-security guardrails apis-json ai-governance prompt-injection llm-security model-safety
-
Updated
Oct 26, 2026
Add this topic to your repo
To associate your repository with the model-safety topic, visit your repo's landing page and select "manage topics."