RENO, Nev., Aug. 11, 2026 (GLOBE NEWSWIRE) -- Gregory Kurtzer, open source founder of Rocky Linux, CentOS, Warewulf, and Apptainer, today launched OpenWALDO, an open source community project building a shared, governed corpus of AI training data, sponsored by CIQ, freeing every builder to innovate and compete above a shared foundation instead of duplicating it.

The launch draws a line the AI industry has been blurring for years. Open weight models, which let anyone download and run a model, have been a valuable milestone, giving developers direct access to powerful systems. Open source models go a step further: the code, the methods, and the training data behind them are open too, so anyone can inspect, verify, and leverage that content so that everyone levels up. An AI Bill of Materials (BOM) builds on that same transparency, giving AI models a verifiable record and the trust that comes with it. OpenWALDO puts that fuller standard into practice, starting with the piece that has stayed closed the longest, the training data itself.

Open source communities have already given the world its computing foundation: Linux, cloud infrastructure, containers, supercomputing, Python, PyTorch, and much of today's AI ecosystem, but that leaves out a critical section of the puzzle, the AI data sources. OpenWALDO is built to provide that foundational open source layer for the entire AI community.

“I’ve spent my career watching open source turn users into builders, competitors into collaborators, and shared problems into common infrastructure that operates at massive scale. No single organization could build or sustain all of that alone. OpenWALDO brings that proven model to AI. Let’s work together, build its foundation in the open, and collaboratively take AI to the next level,” Kurtzer said.

That foundation is missing for a specific reason. Increasingly, AI models are trained on data pulled from many different sources, including other AI models. A model answers a question, and that answer becomes training material for the next model. Researchers call what happens next model collapse, the same way a copy of a copy loses fidelity with every generation. Compounding that, there is often no way to know what data trained a given model, under what license, or with what consent. Does the model include unlicensed copyrighted material, or copyleft code that could taint an organization’s software stack? OpenWALDO addresses both problems at the source, making training data itself open, licensed, and traceable through an AI BOM that connects sources, licenses, corpus selections, training runs, and model releases in one verifiable chain, giving people and companies the opportunity to collaborate on, add to, and verify the corpus directly.

Today, every AI team assembles its own foundational training material from scratch, duplicating work across the industry. OpenWALDO gives them a common foundation instead: one shared public corpus that the community builds and improves together. A source added once can support many models. A single correction strengthens the record for everyone downstream. Better tooling makes each future contribution easier than the last.

OpenWALDO is a project that AI labs, model providers, and open weight models can build on. A lab or company can take the corpus and its bill of materials as a verified baseline, add its own proprietary data, build, and ship, with a clear, auditable line back to its sources. The foundational work gets done collaboratively, in the open, so every builder, from the largest labs to independent researchers and home labbers, stays free to focus on what they add above it.

The project is open to anyone. Individuals, hobbyists, researchers, institutions, and companies can learn together, contribute their expertise, and build from the same public foundation. As the shared corpus absorbs the duplicated foundational work, innovation moves higher up the stack, and every builder stays free to select data, develop models, differentiate, and compete above it.

Accountability grows with the community. Public review, attributable contributions, Developer Certificate of Origin sign-off, and preserved provenance keep responsibility attached from ingestion through release. The result is a durable public AI commons: community governed, independently verifiable, and available for the entire open source AI community to improve, learn from, and build upon.

CIQ has backed open source infrastructure since Kurtzer founded the company in 2020, supporting Rocky Linux, Apptainer, Warewulf, and Ascender. OpenWALDO extends that commitment to AI's foundational layer, following a true open source model where a company funds and supports the work, but the community governs it and takes the contributions. The index, the toolchain, and the models belong to the community, license clean and forkable, as open source is meant to be.

"Open source has won this argument before," Kurtzer continues. "Proprietary vendors made identical claims about open code: insecure, unaccountable, impossible to trust. They lost, because source, licenses, and provenance made trust apparent. Linux didn't win by being certified safe. It won by being inspectable, forkable, and community validated. AI is missing that same property, and OpenWALDO is how we build it."

OpenWALDO builds the substrate, not the models: the licensed, provenance-tracked corpus and the toolchain that turns it into auditable models, for the labs, companies, researchers, and communities doing the training.

Anyone interested in contributing can join the OpenWALDO community on Slack and GitHub, and participate directly in growing the public corpus.

About OpenWALDO

OpenWALDO (Open Weights, Artifacts, Licenses, Data, Origins) is an open source project building a shared, auditable corpus of AI training data and the toolchain that turns it into verifiable models. OpenWALDO is founder-led and community governed. Gregory M. Kurtzer serves as founder; CIQ backs the project as the sponsor. Sources, licenses, and provenance travel with the data from ingestion through model release, so training data behaves like open source input: named, reviewable, versioned, attributable, and verifiable. Learn more at openwaldo.org.

About CIQ

CIQ delivers secure and performant software infrastructure for the demands of all modern workloads, from the most mundane to the most extreme HPC and AI jobs. Founded in April 2020 by Gregory Kurtzer, the original creator of CentOS, Warewulf, and Singularity (now Apptainer), CIQ is the founding commercial sponsor of Rocky Linux and builds the full Enterprise Linux, HPC, and AI infrastructure stack, from the kernel to the orchestrator. CIQ is trusted by 90% of the world's largest enterprises, with millions of actively deployed Rocky Linux instances worldwide. Learn more at ciq.com.

Media Contact

Lindsay Aamodt

[email protected]

Image: https://www.globenewswire.com/newsroom/ti?nf=OTgwNzM4NiM3NzY5MTM4IzIzMjM4NzI=

Image: https://ml.globenewswire.com/media/MTExNmFmYmEtY2U0YS00ZDQ2LTg2MzEtNTg5Yzc0ZjNiNTQ3LTEzMzU0MjItMjAyNi0wOC0xMS1lbg==/tiny/CIQ.png

Source: CIQ