We wholeheartedly support open source.

On 24 July 2026, a coalition including NVIDIA, Microsoft, Meta, Hugging Face, IBM, Mozilla, Mistral, the Linux Foundation, Y Combinator and others published Open Weights and American AI Leadership. The letter argues that open-weight models — models anyone can download, inspect, modify, and run — are essential to a competitive, broadly shared AI economy.

LeemerLabs agrees. Wholeheartedly. We are an Irish AI lab, based in Waterford, and we work closely with American partners — researchers, model providers, hardware vendors, and open-source communities — whose open weights and tooling we build on every day. Our work depends on those open foundations: open-weight base models, open evaluation practice, documented datasets, and the freedom to adapt systems for European data residency and lower-resource languages such as Irish.

Why this letter matters to us

The letter draws the right historical parallel. Open-source software did not weaken the internet; it became the substrate under it. Open weights are the same kind of bet for AI: a shared foundation on which startups, universities, public bodies, and independent labs can build without asking permission for every experiment.

Premature restrictions on open models would not make the world safer by default. They would concentrate capability, raise costs, and push adaptation work offshore or underground. The letter is clear on this. So are we.

What we endorse

We specifically endorse the letter's core claims:

• Open weights expand access to the AI economy for organisations that cannot — and should not have to — train every model from scratch.
• Competition across models, chips, clouds, and applications keeps the gains of AI from concentrating in a few hands.
• Customers deserve control over data, deployment location, and the value they create when they specialise a model.
• Cybersecurity defenders need models comparable to those available to attackers; openness broadens defensive capability.
• Distillation and related techniques are legitimate tools for improvement, evaluation, and validation. Unlawful extraction from closed systems should be handled with targeted legal and commercial remedies — not sweeping bans that freeze legitimate research.

How we practise this

This is not a press-release posture. It is how LeemerLabs operates:

• We host and serve open-weight models on Irish and EU infrastructure behind an OpenAI-compatible gateway.
• Born, our applied intelligence programme, adapts, evaluates, and ships models with provenance and proof — not marketing fog.
• We treat open source as language infrastructure for Irish and other lower-resource languages: inspectable, adaptable, and shareable.
• We publish methods and release artefacts when we claim progress. Claims follow evidence.

American leadership in open weights helps a plural frontier. So does European capacity to run, harden, and specialise those models under local law. The two are complementary. A closed stack in either place is a single point of failure for everyone else.

Our ask

Policymakers should expand compute access for startups and researchers, invest in shared datasets and evaluation frameworks, and avoid premature restrictions that stifle competition or drive open innovation elsewhere. Keep the frontier plural. Keep the substrate open.

LeemerLabs stands with the signatories of aka.ms/OpenLetter. We wholeheartedly support open source. We will keep building in the open.