Open Data and Models
Data is the heart of AI. And models are the heart of the intelligence of an AI system. The AI Alliance includes most of the companies and research labs that have built and released open models. The AI Alliance is focused on studying, tuning, adapting and applying open models and releasing them to advance open science and research and ensure they are useful to many people and organizations.
TAPESTRY
The foundational layer of the Open Planetary AI Network — open, shared AI infrastructure where any nation or industry can build AI in their own language, culture, and data, without depending on closed infrastructure they don't own. Led by Yann LeCun as Chief Scientific Advisor.
Open Trusted Data Initiative (OTDI)
A catalog of open datasets scored by trust, built on a metadata specification for data provenance, lineage, and utility. AI systems are only trustworthy if the data is trustworthy.
SYNTH
An open, synthetic, multilingual data corpus for training sovereign AI models and advanced agents, built to close a gap in open-source AI development.
Semikong
A fully open foundation model for the semiconductor industry — the template for domain-specific open models that compete with proprietary alternatives.
The Open Agent Hub
The Open Agent Hub is a collaborative community of open source AI projects and domain-specific work groups that seek to make AI Agents successful in the real world through fast experimentation and distillation of learning into usable reference architectures and implementations, and build out of new tools to enable development and deployment.
Semiont
A wiki-like knowledge base with graph retrieval, where humans and AI agents co-create knowledge. Use it as a wiki, a knowledge base, or agent memory.
Dana Agent Framework
An open framework for building reliable, domain-aware agents that connect language models to real-world tools and workflows.
Llama Stack
Standardizes the core building blocks of AI application development and integrates with open source tools and managed services.
Open Agent Hub
A community hub of open source blueprints, evaluation frameworks, and domain-specific architectures for production AI agents. Agents fail at "demo to deployment" — domain-specific architecture fills the gap.
CUBE
The Common Unified Benchmark Environment: a standard for wrapping benchmarks, a harness for building and evaluating AI agents, and a community registry of compatible benchmarks.
AllyCat
A beginner-friendly, end-to-end RAG application that crawls a website and lets you ask questions of its content using open models.
Gofannon
A web app for prototyping AI agents and their user interfaces, with a library of functions other agent frameworks can use.
Deep Research Agent for Applications
Example applications that use MCP to implement the deep research agent pattern, with working examples for finance, medicine, and arXiv exploration.
MCP in the Enterprise: A User Guide
A living guide, written by practitioners, to deploying, managing, and securing MCP and related agent protocols in enterprise settings.
Safety and Governance
AI Alliance's safety, governance, and advocacy work builds public trust in open AI through evaluation, benchmarking, and red-teaming, alongside advocacy for healthy, sustainable, and open regulatory environments and technical standards across the industry.
Testing Generative AI Agent Applications
Creates reusable techniques and examples for deterministic and generative AI evaluation, helping enterprises ensure deployed AI meets safety and requirements standards.
Evaluation Is for Everyone
Designed for all AI builders, this initiative offers accessible trust and safety evaluation tools, making robust evaluations easy to find and implement for any AI application.
Evaluation Reference Stack
This companion project aggregates flexible, easy-to-use tools and packages for writing and executing AI evaluations, serving as a runtime stack for broader trust and safety initiatives.
