Run Claude Code 100% on-device with local AI on Apple Silicon. MLX-native Anthropic-API server, 65 tok/s Qwen 3.5 122B, Llama 3.3 70B, Gemma 4 31B. Private, offline, airgap-ready. Built for NDA / legal / healthcare workflows.

Powerful search page powered by LLMs and SearXNG

A deterministic privacy boundary between your data and AI.

A curated list of academic events on AI Security & Privacy

When the stakes are high, intelligence is only half the equation - reliability is the other ⚠️

[ICML 2024 Spotlight] Differentially Private Synthetic Data via Foundation Model APIs 2: Text

Convert Word docs to Markdown privately - 100% offline, no uploads. Perfect for processing sensitive documents with Ollama, LM Studio, GPT4All & other local AI tools. Just double-click the standalone/word-to-markdown.html file to use.

International secret & PII redaction for JS/TS: logs, prompts, HTTP, datasets. 20+ languages, checksum-validated national IDs, reversible LLM layer. Zero deps, ReDoS-safe.

OfflineAI is an artificial intelligence that operates offline and uses machine learning to perform various tasks based on the code provided. It is built using two powerful AI models by Mistral AI.

Comprehensive taxonomy of AI security vulnerabilities, LLM adversarial attacks, prompt injection techniques, and machine learning security research. Covers 71+ attack vectors including model poisoning, agentic AI exploits, and privacy breaches.

Local RAG system with a built-in governance agent that filters sensitive or restricted information with separated agent logging systems to keep privacy and security

Automatically redacts sensitive data in screenshots before sending to AI agents

[ICCV 2025] Geminio is a VLM-powered gradient inversion attack in federated learning (FL). It allows the adversary (the FL server) to describe the data of value and reconstruct the victim client's private data matching the description.

A practical guide to AI privacy, profiling, shadow profiling, local AI, cloud AI, and the future of human autonomy.

The LLM Unlearning repository is an open-source project dedicated to the concept of unlearning in Large Language Models (LLMs). It aims to address concerns about data privacy and ethical AI by exploring and implementing unlearning techniques that allow models to forget unwanted or sensitive data. This ensures that AI models comply with privacy.

awesome list of multi-agent security resources

A transparent, local-only tool to sanitize sensitive info for AI.

Local de-identification proxy for AI coding agents. Mask secrets and PII before Claude Code or Codex traffic leaves your machine.

Semantic PII Masking & Anonymization for LLMs (RAG). GDPR-compliant, reversible, and context-aware. Supports LangChain & OpenAI

🔒 Detect security leaks in AI-assisted codebases. Static analysis tool for Python & JS/TS with cross-file taint tracking.