Twenty companies and organizations have signed an open letter calling on U.S. policymakers to protect promiscuous AI models.
The letter (PDF) released today features signatures from a list spanning direct commercial rivals and organizations with little apparent overlap in business models, including Meta, Microsoft, Nvidia, IBM, Dell Technologies, CrowdStrike, Palantir, ServiceNow, Hugging Face, Perplexity, Mistral, Andreessen Horowitz, Y Combinator, Linux Foundation, and Mozilla.
The letter’s discussion focuses on a comparison between the open source software movement of the 1980s and the current battle over whether the weight of AI models should circulate freely or remain locked behind commercial APIs.
Openweight models are AI systems whose trained parameters are public and anyone can download, inspect, modify, and run them on their own hardware. This differs from closed models where the underlying weight never leaves the vendor’s infrastructure, such as Frontier products offered by OpenAI and Anthropic through API access only.
The signatories frame Openweight as a mechanism by which AI capabilities extend beyond a few well-funded labs and into what the letter calls the workflows of “factories, hospitals, farms, classrooms, and high street businesses.”
Their argument unfolds in three tracks.
Openweight reduces the cost of entry for startups and public institutions that cannot afford to train Frontier models from scratch or pay per-token fees at Frontier prices for routine tasks. Openweight will foster competition across the stack, from chips to cloud infrastructure to applications, which will keep costs down and prevent value from being captured by a few providers, the letter said. OpenWeight also provides enterprise customers with a way to avoid vendor lock-in, as organizations running the OpenWeight model can manage their own data and adapt the model to their internal requirements without relying on a single service. Vendor roadmap or pricing.
Discussions about security go against instinct.
The most incisive sections of the letter address the risk case directly, reversing the usual framework around open models and security, and are therefore well worth reading.
Once weights are released, they are beyond the control of the original developer, the letter acknowledges. It becomes difficult to track and revert modified versions. A tweaked or bare-bones version of the open model can be cycled with the safety guardrail removed and has no recall mechanism.
The signatories argue that the answer is not prohibition. Their argument is based on a comparison with cybersecurity. Defenders facing AI-powered attackers need access to models with comparable capabilities to detect and simulate threats, something that closed, permission-gated systems cannot easily provide.
They extend this to broader security claims, arguing that closed models are inherently insecure because they can be compromised, exploited, or fail in ways that cannot be observed or verified by outside researchers. After reading this article, you’ll find that concentrating advanced functionality behind a small number of private providers creates a single point of failure rather than eliminating one.
In contrast, an open model allows external researchers to investigate behavior, perform red team exercises, and identify vulnerabilities across many teams, rather than relying on one vendor’s internal testing.
The letter draws direct parallels to the “open source is safer than unknown” argument that has shaped the software security debate for decades, but it does not cite specific vulnerability discovery data or incident numbers to support this claim, especially as it applies to AI systems.
Distillation gains certain protection
The letter opens up a space for one of the more controversial techniques in AI circles: distillation, where the output of one model is used to train or improve a second model. It is a standard technique in machine learning research and product development, used to evaluate, validate, and migrate functionality between models of different sizes.
The signatories draw a line between legal distillation and what they call “illegal efforts to extract value from closed models,” arguing that the former should not be caught up in restrictions aimed at the latter.
This is a direct response to the controversy that erupted after the rise of Chinese models such as DeepSeek and Kim. That’s when several US laboratories suggested that competing models were being trained by unauthorized distillation of the output from their own closed systems.
The position of this letter is to address abuse through targeted legal and commercial mechanisms, rather than blanket restrictions on technology that the entire sector relies on.
What does this imply for future policy battles?
No specific legislative or regulatory proposals are attached to this letter. It is a positioning document ahead of expected AI policy action in Washington, urging lawmakers to expand access to computing for startups and researchers, fund shared training datasets and evaluation frameworks, and avoid so-called “premature restrictions” on open models.
This should be treated as an indicator of where major infrastructure and chip providers want the regulatory debate to land, rather than as a result of settled policy. Companies like Nvidia, IBM, and Dell have direct commercial reasons to want the open weight ecosystem to flourish, as it expands the range of models they can deploy and sells more compute and services, regardless of the lab that produced the weights.
Procurement teams weighing open-weight versus closed-model deployments should consider that the policy environment favoring one approach over the other remains unresolved, and that restrictions on distillation or open release could change the economics of self-hosted AI within a single legislative cycle.
See: OpenAI pushes ChatGPT to patient health records
Want to learn more about AI and big data from industry leaders? Check out the AI & Big Data Expos in Amsterdam, California, and London. This comprehensive event is part of TechEx and co-located with other major technology events such as Cyber Security & Cloud Expo. Click here for more information.
AI News is brought to you by TechForge Media. Learn about other upcoming enterprise technology events and webinars.
