Kurtzer's Next Gambit: OpenWALDO Aims to Be the Linux for AI Data

RENO, Nev. – August 11, 2026

Gregory Kurtzer, a figure synonymous with major shifts in the open source landscape through projects like CentOS and Rocky Linux, has once again thrown down the gauntlet. This time, the target isn't the enterprise operating system, but the very foundation of artificial intelligence: its training data. With the launch of OpenWALDO, Kurtzer and his company CIQ are making a strategic wager that the principles of transparency, community governance, and verifiability that won the software wars can solve AI's most pressing and expensive problems.

OpenWALDO (Open Weights, Artifacts, Licenses, Data, Origins) is an open source community project with a deceptively simple and profoundly ambitious goal: to build a shared, governed, and completely auditable corpus of AI training data. This isn't merely another repository of datasets. It's a foundational maneuver designed to address the industry's growing crises of trust, transparency, and sustainability. As AI models become more powerful, the data they are trained on has remained stubbornly opaque, creating a house of cards built on unknown sources, questionable licenses, and unverifiable claims. Kurtzer's move signals a belief that the only way to build a skyscraper is to first collaboratively engineer the bedrock.

The Transparency Imperative in a 'Collapsing' AI World

The AI industry is grappling with a paradox. While capabilities are soaring, transparency is in decline. A recent Foundation Model Transparency Index showed a troubling trend toward opacity among major AI labs. This lack of insight isn't just an academic concern; it's a ticking time bomb for enterprise adoption. Without knowing what data a model was trained on, companies face immense risks from copyright infringement, privacy violations, and embedded biases that can surface at the worst possible moments.

Compounding this is the looming threat of 'model collapse.' Researchers have warned that as AI models are increasingly trained on the synthetic data generated by other models, they begin to lose fidelity, like a photocopy of a photocopy. The system starts feeding on itself, forgetting the nuances of original human-generated data and amplifying its own flaws. This digital inbreeding threatens to cap the potential for future innovation. OpenWALDO attacks this problem at its source by providing a stable, high-quality baseline of foundational data that is verifiable back to its human origins.

Currently, every AI team, from garage-based startups to mega-corporations, largely duplicates the same foundational work: scraping the web, cleaning datasets, and wrestling with a chaotic tangle of licenses. This represents a colossal waste of resources. OpenWALDO proposes a new model. "Let’s work together, build its foundation in the open, and collaboratively take AI to the next level," Kurtzer stated in the launch announcement. The project aims to create a common infrastructure layer, allowing developers to stop rebuilding the same foundation and instead focus their unique value higher up the stack.

An AI Bill of Materials: Building the Verifiable Chain

The core mechanism behind OpenWALDO's vision is the AI Bill of Materials (BOM). The concept extends the software BOM—a list of components in a piece of software—to the entire AI lifecycle. For the first time, this provides a structured, machine-readable inventory that connects every piece of the puzzle: the original data sources, their licenses, the corpus selections used for a specific training run, and the final model release. It creates a verifiable chain of provenance.

This isn't just metadata; it's an active system of governance. Using established open source tools, the project ensures accountability. Git is used to govern the meaning of metadata, making changes reviewable and attributable. Canonical data objects are content-addressed with hashes, giving them a unique identity. Contributor sign-offs ensure responsibility is attached to every addition. The result is a system where training data behaves like well-managed open source code: named, reviewable, versioned, and auditable. An AI lab can take this verified baseline, add its own proprietary data for differentiation, and still maintain a clear, auditable line back to its sources.

This approach draws a sharp distinction between 'open weight' and true 'open source' AI. An open weight model, while useful, is often a compiled binary; you can run it, but you can't inspect its ingredients or rebuild it from scratch. OpenWALDO aims to provide the full recipe, enabling anyone to inspect, verify, and even fork the entire process. As Kurtzer puts it, "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."

CIQ's Strategic Play for the AI Foundation

The sponsorship of OpenWALDO by CIQ, the company Kurtzer founded in 2020, is a critical piece of this strategic puzzle. CIQ has built its business on providing enterprise-grade infrastructure and support for a full stack of open source technologies, from its flagship Rocky Linux to the Apptainer container system and Fuzzball HPC orchestration. The company's move to back a foundational data layer for AI is a natural and potent extension of this strategy.

By funding OpenWALDO, CIQ is positioning itself at the center of the next wave of AI infrastructure. The maneuver is clear: if OpenWALDO becomes the trusted, standard commons for AI training data, then CIQ, with its deep expertise in the project's architecture, is perfectly placed to offer the commercial support, integration, and enterprise-grade tooling that large organizations will require to build upon it. This mirrors the successful business model that has sustained enterprise open source for decades: foster a vibrant, community-governed project while providing the paid services that bridge it to corporate needs.

This move also serves to unify CIQ's product portfolio into a cohesive AI narrative. An enterprise can now potentially rely on a CIQ-supported stack that runs from the base operating system (Rocky Linux) and containerization (Apptainer) all the way to the foundational data itself (OpenWALDO). It's a bold play to become the go-to provider for organizations that want to build AI systems on a transparent, secure, and fully open source foundation.

The Open Source Gauntlet: Challenges and the Path Forward

Despite the compelling vision, the path for OpenWALDO is fraught with challenges. The legal landscape surrounding AI training data is a minefield of copyright lawsuits and evolving privacy regulations like the GDPR and the EU AI Act. Creating a truly 'license-clean' corpus at scale will require immense diligence and a robust governance model to vet contributions. Furthermore, the project will have to contend with the monumental task of mitigating inherent biases within vast datasets, a problem that transparency alone does not solve.

Attracting a critical mass of diverse contributors—from individual hobbyists to academic institutions and competing corporations—will be paramount. The project's success hinges on the network effect; its value grows exponentially with each high-quality contribution. It must prove that its model is more efficient and reliable than the siloed efforts of individual companies or the less-structured collections of other data hubs.

Ultimately, Gregory Kurtzer is betting that the open source playbook is universal. He's arguing that the same forces of collaborative development and radical transparency that commoditized the proprietary operating system can now establish a new, trusted foundation for artificial intelligence. If he's right, OpenWALDO won't just be another project; it will be the substrate upon which the next generation of trustworthy and innovative AI is built.