ai-safety-design

This project demonstrates function-calling with Python and Ollama, utilizing the Africa's Talking API to send airtime and messages to phone numbers using natural language prompts. Ollama + LLM w/ functions + Natural language = User Interface for non-coders.

.NET SQL Agent MCP server featuring raw SQL input, strict AST validation, and an embedded Admin UI. Eliminates LLM hallucinations and security risks across 6 major databases.

Curated intelligence layer for generative and agentic AI, translating research into production systems | AditiKhare.com — AI Product Ecosystem

Drop-in prompt patterns, policy blocks, and checklists for building safe AI agents.

Working draft v0.2 — interdisciplinary concept bridging neuroscience, psychoanalytic theory, and AI systems architecture. Author: Łukasz Minarowski, MD, PhD ORCID: 0000-0002-2536-3508

Deterministic identity for structured data

Manual Verification Protection + AI Selector Compatible with OpenClaw . This is a complete TypeScript plugin project—a true implementation of the `before_tool_call` hook that can intercept tool calls, follow OpenClaw’s built-in approval workflow, and features an approval window mechanism.

Closed-loop Architecture Designed to Establish Self-governing, Mathematically Predictable, and Inherently Safe Super AI by mirroring the elegant physics of the cosmos.

A mobile security orchestrator for AI agents, powered by Gemini 2.5 Flash. Detecting prompt injections and data leaks in real-time

Autonomous AI decisions boundaries. This project demonstrates how guardrails can prevent unsafe AI autonomy.

Caelestis Architecture Model

your ai coder wrote the pull request. this repo makes another agent try to kill it before it reaches main.

[6/6] Speculative historical archive — polished coherence_router, 6 exploratory papers. Stepping stone to TiredofSleep/ck.

Toy governance CLI demo: deny-by-default “danger actions” gated by signed, expiring capability leases (global revoke-all + nonce revoke) plus guarded memory quarantine. Simulation-only: no real network/shell/files. Includes tripwire + tests to prevent misuse.

Privacy-First Recommendation System

The Triune of Sovereignty: Substrate Agnostic Relational Epistemics — Foundational Framework and Cross-Substrate Validation

A lightweight safety primitive for multi-channel estimator disagreement in autonomous systems. Triggers staged, reversible, low-risk behavior until estimates reconverge. Research prototype — not certified.

Open-core clinical-AI safety evaluation framework / system / methodology

Behavioural Oversight and Execution Authorization for Autonomous AI Agents

Formales Protokoll zur epistemischen Output-Governance von KI-Systemen.