LONDON, 28 July 2026 — Scrums.com, the Software Engineering Orchestration Platform signs support for open models and the Open Weights Letter published by Nvidia and Microsoft which already has approximately 140 signatories, including NVIDIA, Meta, Google, Hugging Face, Palantir, SpaceX, Vercel and the Linux Foundation, and makes the case for a competitive, secure and widely accessible open-weight AI ecosystem.
Scrums.com builds and operates AI systems for some of the largest global companies in finance, technology, and the consumer internet. These production environments require control over data, infrastructure and model selection. Scrums.com cited those customer requirements, together with evidence from production deployments, as the basis for supporting the letter.
Public discussion about AI capability often focuses on the AI model itself. In production, the surrounding harness architecture matters just as much. The harness gives an agent context, tools, authority guardrails, stability and evaluations to measure performance.
As the Linux Foundation explains:
“The harness is the scaffolding that determines how an agent behaves in practice.”
Scrums.com's Software Engineering Orchestration Platform (SEOP) provides that agent harness for enterprise software delivery. It coordinates models, agents, tools, infrastructure, permissions and evaluation across the delivery lifecycle.
Once the harness determines how the system behaves, which model to use becomes a commercial decision. Models can be benchmarked for each task and selected according to capability, cost, latency, security and data-residency requirements. That approach depends on organisations having access to a broad and competitive global market of large language models (LLMs).
The Open Weights Letter connects this to lock-in, and to the risk that organisations can lose the AI capability they build over time:
“Open weight models help provide that assurance by allowing organizations to control their own data, evaluate and adapt models to their own needs, and deploy them wherever their business requirements demand.”
For enterprise buyers, this is a practical procurement requirement. Questions about data jurisdiction, provider dependency and ownership of accumulated knowledge recur across regulated industries. In many cases, open-weight models provide the clearest route to meeting those requirements.
AI systems built and operated by Scrums.com are model-agnostic. Customers can change models as requirements evolve and deploy systems within their own cloud environment, private infrastructure or customer-controlled data centre.
“The AI models make the headlines, but the harness makes them workable in the enterprise. Our Scrums.com customers don't yet ask us about open weights. They ask who controls their data and their IP, and whether they can move off an LLM provider later. Open weights are the only long term answer.”
Gerald Neves, CEO of Scrums.com
The letter also recognises the risks of open weights. Once released, model weights cannot easily be withdrawn, and modified versions can be difficult to trace. Scrums.com supports the letter's position that these risks should be addressed through stronger security practices, transparent evaluation and a more capable open global ecosystem rather than broad restrictions on open-model development.
The full original letter is embedded below.
Open Weights and American AI Leadership