A system that turns jailbreak papers into runnable attacks and benchmarks — live, as research evolves.
-
Updated
Jul 15, 2026 - Python
#
ai-safety-research
Here are 95 public repositories matching this topic...
An ongoing, collaborative meta-analysis about Human-AI-Interactions. We aggregate data and knowledge to build a non-abrasive, user-friendly prompting framework tailored to LLM mechanics, ensuring reasoning stability and a friction-free prompting environment that is safe for the human psyche and wellbeing.
open-science alignment meta-analysis call-to-action prompt-engineering chain-of-thought sycophancy agentic-ai llm-behavior ai-safety-research anti-sycophancy
-
Updated
Aug 25, 2026 - HTML
[ICML 2026] AutoControl Arena: Frontier AI Risk Auto-Discovery Platform
-
Updated
May 2, 2026 - Python
200 AI agent skills, hardened with targeted behavioral guardrails. Free drop-in replacements.
mcp ai-safety ai-agents ai-security ai-skills ai-safety-research ai-behavior-analysis ai-skill-safety
-
Updated
Apr 21, 2026 - JavaScript
The left hemisphere. Frameworks, logic, and certainty architecture. Home of FSVE, AION, LAV, ASL, GENESIS, TOPOS, and 60+ epistemically validated frameworks built to make AI systems reliable, not just capable.
agi ai-safety systems-architecture medical-ai ethical-ai ai-policy ai-frameworks trustworthy-ai ai-governance agentic-ai ai-assessment ai-auditing ai-architect epistemic-ai ai-safety-research legal-ai-safety financial-ai-safety safety-invariants sheldon-k-salmon full-dive
-
Updated
Apr 15, 2026 - Python
A minimal decoder probing GPT-2's residual stream activations for mechanistic interpretability research .
-
Updated
Jul 1, 2026 - Python
A commit-and-audit proof system for deterministic, quantized inference of a JEPA-style world model (LeWorldModel)
-
Updated
Jun 12, 2026 - Rust
AgenticStore: The secure toolkit for AI agents. Instantly equip Claude Desktop, Cursor, and Windsurf with 27+ MCP tools, persistent memory, and SearXNG search, all protected by a built-in PII prompt firewall to protect your data from being exposed to AI agents.
open-source mcp ai-agents ai-security ai-tools prompt-tuning ai-governance ai-tools-directory antigravity model-context-protocol mcp-server ai-skills claude-code model-context-protocol-servers trending-ai ai-safety-research
-
Updated
Apr 9, 2026 - Python
The SAPIEN Framework — an open standard (CC BY 4.0) for measuring AI behavioral safety (sycophantic drift), plus voigt-kampff, the FSL scoring CLI.
benchmark alignment red-team ai-safety blue-team drift-detection behavioral-safety ai-red-team llm-evaluation llm-safety ai-safety-research sycophancy-detection ai-blue-team
-
Updated
Aug 8, 2026 - Python
I am investigating the mechanistic architecture of unfaithful Chain-of-Thought (CoT), specifically mapping and disrupting the “shortcut circuits” that allow models to bypass explicit reasoning.
-
Updated
Jul 6, 2026 - Jupyter Notebook
👟 SUP: Sycophancy Under Pressure
-
Updated
Jan 11, 2026 - Python
CoverAgent is a targeted behavioral evaluation harness designed to red-team multi-agent systems for collusion
-
Updated
Jul 27, 2026 - Python
AI security scanner for OpenClaw - powered by AgentTinman. Discovers prompt injection, tool exfil, context bleed, and other security issues in your AI assistant sessions, then proposes mitigations mapped to OpenClaw's security controls.
-
Updated
Feb 17, 2026 - Python
A testbed for the Animal Harm Benchmark.
-
Updated
Mar 1, 2026 - Python
Closed-loop Architecture Designed to Establish Self-governing, Mathematically Predictable, and Inherently Safe Super AI by mirroring the elegant physics of the cosmos.
agi ai-safety systems-engineering cognitive-architecture ai-alignment cognitive-architectures superintelligence artificial-superintelligence systems-engineering-ai ai-safety-design ai-safety-research ai-alignment-practice ai-alignment-research ai-safety-philosophy
-
Updated
Jul 22, 2026
Grid environment for studying intelligent disobedience in cooperative leader-follower multi-agent systems (AAMAS '26)
-
Updated
Jul 28, 2026 - Python
A kernel-userland protocol enforcing information-theoretic bounds on AI adaptivity leakage, benchmark gaming, and capability spillover.
rust cryptography kernel webassembly wasm verification sandboxing formal-methods systems-programming ai-safety differential-privacy reference-monitor e-values agentic-ai deterministic-execution ai-safety-research epistemic-security benchmark-integrity
-
Updated
Mar 5, 2026 - Rust
Neuro-symbolic framework for autonomous remediation of cloud infrastructure misconfigurations using Z3 SMT verification and LLM patch generation
benchmark terraform cloud-infrastructure z3 infrastructure-as-code smt-solver formal-verification ai-safety cloud-security neuro-symbolic llm llm-benchmark autonomous-remediation ai-safety-research cloudfix-bench
-
Updated
Aug 13, 2026 - Python
Shield models AI safety the way humans experience safety
-
Updated
Sep 3, 2026 - Go
To Learn Without the Possibility of Undoing is not Intelligence, It's a Surrender to Emergence.
experimentation research-paper security-research robust-machine-learning svar low-rank-adaptation ai-safety-research rlae reversible-behavoural-learning
-
Updated
Jun 8, 2026 - Jupyter Notebook
Add this topic to your repo
To associate your repository with the ai-safety-research topic, visit your repo's landing page and select "manage topics."