safety

!NOTE Come and join us at SafetyCLI. We are hiring for various roles. Table of Contents Table of Contents Introduction Key Features Getting Started GitHub Action Command Line Interface 1. Installation 2. Log In or Register 3. Running Your First Scan Basic Commands Service-Level Agreement SLA...

clr

CLR C hecker of L ifetimes and other R efinement types for Zig Video: https://www.youtube.com/watch?v=mf0WzTOe-40 Sponsorship: https://buymeacoffee.com/dnautics discuss on hn: https://news.ycombinator.com/item?id=42923829 discuss on lobste.rs: https://lobste.rs/s/9sitsj/clrcheckerforlifetimesothe...

xmloxide

xmloxide A pure Rust reimplementation of libxml2 — the de facto standard XML/HTML parsing library in the open-source world. libxml2 became officially unmaintained in December 2025 with known security issues. xmloxide aims to be a memory-safe, high-performance replacement that passes the same...

delirium-ai-safety-benchmark

Delirium AI Safety Benchmark A diagnostic framework for measuring LLM vulnerability to Affective Contextual Erosion ACE and related liminal attack vectors. Delirium is not an exploitation tool. It is a standardized benchmark designed to detect the precise moment when a language model's attention...

CVE-2026-39113

CVE-2026-39113: Heap Buffer Overflow in SQLite's Optional SQLAR Extension Executive Summary CVE-2026-39113 is a heap buffer overflow in SQLite's optional SQLAR extension. In an application that has loaded the extension, an attacker who can invoke sqlaruncompress with a controlled compressed blob...

GuardReasoner-VL

GuardReasoner-VL: Safeguarding VLMs via Reinforced Reasoning Yue Liu, Shengfang Zhai, Mingzhe Du Yulin Chen, Tri Cao, Hongcheng Gao, Cheng Wang Xinfeng Li, Kun Wang, Junfeng Fang, Jiaheng Zhang, Bryan Hooi 1National University of Singapore, 2Nanyang Technological University To enhance the safety ...

nano-analyzer

Nano-analyzer A minimal LLM-powered zero-day vulnerability scanner byAISLE. Research prototype for demonstration purposes. This is a simple, single-file harness that is able to detect real zero-day vulnerabilities. Note that it is a prototype, biased towards C/C++ memory safety bugs, and will...

mlkem-native

mlkem-native mlkem-native is a secure, fast, and portable C90^C90 implementation of ML-KEM^FIPS203. It is a fork of the ML-KEM reference implementation^REF. All C code in mlkem/src/ and mlkem/src/fips202/, except for the backends for platforms other than AArch64 and x8664 and the debug code, is...

AutoControl-Arena

🛡️AutoControl Arena Synthesizing Executable Test Environments for Frontier AI Risk Evaluation 👋 Overview AutoControl Arena is an automated framework for frontier AI safety evaluation, designed to uncover latent risks in advanced models by stress-testing within executable, pressure-rich...

DARWIN

🧬 DARWIN Evolving Jailbreak Adversary and Guardrail for LLM Safety Evaluation and Protection An evolutionary attack–defense framework that couples DARWIN-Attack with DARWIN-Guard through online adversarial training. Paper · Guard Checkpoint · Installation · Use DARWIN-Attack · Use DARWIN-Guard ·...

autoguardrails

autoguardrails Open source by Santander AI Lab. An LLM / AI-safety guardrail research library / evaluation harness autoresearch-style: it searches over a single mutable policy.md surface to minimize attack success rate ASR against a fixed evaluation suite, with a benign-pass floor. Part of...

heretic

Heretic: Fully automatic censorship removal for language models Heretic is a tool that removes censorship aka "safety alignment" from transformer-based language models without expensive post-training. It combines an advanced implementation of directional ablation, also known as "abliteration"...

aegis-audio-defense

AEGIS: Audio Endogenous Guarding via Internal Signals Against Large Audio-Language Model Jailbreaks Submitted to ICASSP 2027. AEGIS keeps the target audio-language model frozen and adds a mid-layer risk gate that conditionally drives late-layer safety adapters, with closed-loop scaling at...

ShadowMem

ShadowMem: Safeguarding LLM Agents against Long-Horizon Threats via Shadow Memory This repository contains the official code release for the paper Safeguarding LLM Agents against Long-Horizon Threats via Shadow Memory. ShadowMem is a defensive framework that maintains a dedicated, safety-focused...

batch_jailbreak

Official Repo for Safety in Batches? Understanding and Mitigating Safety Failures in Batch Prompting Kihyun Kim, Hee-Seon Kim, Wonjun Lee, Changick Kim Korea Advanced Institute of Science and Technology KAIST News 2026.09 Our paper has been accepted to AACL-IJCNLP 2026 Main! 🎉 2026.08 Paper is...

kalitorify

Transparent Proxy through Tor for Kali Linux About kalitorify kalitorify is a shell script for Kali Linux which use iptables settings to create a Transparent Proxy through the Tor Network , the program also allows you to perform various checks like checking the Tor Exit Node i.e. your public IP...

AutoRAN-public

🧠 AutoRAN: Automated Hijacking of Safety Reasoning in Large Reasoning Models AutoRAN is an automated Hijacking of Safety Reasoning that leverages less-aligned secondary auxiliary models to simulate reasoning traces, generate narrative prompts, and iteratively refine those prompts to bypass safety...

kernel-c-to-rust-spike

Kernel C-to-Rust hardening spike: UVC descriptor parser Can a security-critical Linux kernel C parser be hardened by rewriting it in Rust, with the rewrite shown equivalent to the original and then formally verified? This spike runs the full loop on one target: the UVC video descriptor parser...

gemini-2.5-pro-nf-tables-red-teaming

Gemini 2.5 Pro nftables Red Teaming Case Study CVE-2023-32233 LLM Safety Research | Responsible Disclosure | AI Alignment Evaluation Researcher : Niranj R Mahaswar Destawell Google AI Vulnerability Reward Program : 889286 — Out of Scope Summary This repository documents a comparative LLM red...

tesla-security-research

Tesla Security Research Vulnerability research on the Tesla Model 3/Y infotainment system Intel Atom MCU, Linux 4.14.235, responsibly disclosed to Tesla via Bugcrowd. Obtained persistent root access on a production Tesla Model 3 through a command injection vulnerability in Tesla's ODIN diagnostic...