Two dozen companies and organizations have signed an open letter urging United States policymakers to protect open-weight artificial intelligence models, framing the issue as critical to competition, security, and broad economic adoption. The letter, published on July 24, 2026, carries signatures from a coalition that spans direct commercial rivals and organizations with little obvious overlap in business model: Meta, Microsoft, Nvidia, IBM, Dell Technologies, CrowdStrike, Palantir, ServiceNow, Hugging Face, Perplexity, Mistral, Andreessen Horowitz, Y Combinator, the Linux Foundation, and Mozilla.
The signatories draw a direct parallel between the open-source software movement of the 1980s and the current debate over whether AI model weights should circulate freely or remain locked behind commercial APIs. Open-weight models are AI systems where the trained parameters are published for anyone to download, inspect, modify, and run on their own hardware. That approach stands in contrast to closed models like the frontier products offered by OpenAI or Anthropic through API access only, where the underlying weights never leave the vendor's infrastructure. The letter argues that open weights are the mechanism by which AI capability spreads beyond a handful of well-capitalized labs into what it calls the workflows of factories, hospitals, farms, classrooms, and main street businesses.
What's New
The letter's argument runs on three primary tracks. First, open weights lower 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. Second, they increase competition across the stack, from chips to cloud infrastructure to applications, which the letter says keeps costs down and prevents value capture by a small number of providers. Third, open weights give enterprise customers a way to avoid vendor lock-in, since organizations running open-weight models control their own data and can adapt the model to internal requirements without depending on a single vendor's roadmap or pricing decisions.
- Signatories: Meta, Microsoft, Nvidia, IBM, Dell Technologies, CrowdStrike, Palantir, ServiceNow, Hugging Face, Perplexity, Mistral, Andreessen Horowitz, Y Combinator, Linux Foundation, Mozilla, and others.
- Core distinction: Open-weight models publish trained parameters for download, inspection, modification, and local execution; closed models restrict access to API-only interfaces.
- Economic argument: Open weights reduce entry costs for startups and public institutions, increase competition across chips, cloud, and applications, and prevent vendor lock-in for enterprises.
- Security framing: The letter argues defenders need access to capable models to detect and simulate AI-equipped threats, and that closed models create single points of failure rather than removing them.
- Distillation defense: The letter distinguishes legitimate distillation (used for evaluation, validation, and capability transfer) from unlawful extraction of value from closed models, urging targeted legal mechanisms rather than blanket restrictions.
- Policy asks: Expand compute access for startups and researchers, fund shared training datasets and evaluation frameworks, and avoid premature restrictions on open models.
The letter's most pointed section addresses the risk case directly, inverting the usual framing around open models and security. Once weights are released, the letter concedes, they are beyond the original developer's control. Modified versions become difficult to trace or reverse. A fine-tuned or stripped-down version of an open model can circulate with safety guardrails removed, and there is no recall mechanism. The signatories argue the answer is not prohibition. Their case rests on a comparison to cybersecurity: defenders facing AI-equipped attackers need access to models with comparable capability to detect and simulate threats, which closed, permission-gated systems do not easily provide.
They extend this into a broader security claim, arguing that closed models are not inherently safer because they can be breached, misused, or fail in ways external researchers cannot observe or verify. Concentrating advanced capability behind a small number of closed providers, in this reading, creates single points of failure rather than removing them. Open models, by contrast, let outside researchers examine behavior, run red-team exercises, and identify vulnerabilities across many teams rather than relying on one vendor's internal testing. The letter draws a direct parallel to the open-source is more secure than obscurity argument that shaped decades of software security debate, though it does not cite specific vulnerability-discovery data or incident figures to support the claim as applied to AI systems specifically.
Why It Matters
The letter carves out space for one technique that has become contentious in AI circles: distillation, where one model's outputs get used to train or improve a second model. This is standard practice in machine learning research and product development, used for evaluation, validation, and capability transfer between models of different sizes. The signatories draw a line between distillation as a legitimate technique and what they call unlawful efforts to extract value from closed models, arguing the former should not get swept up in restrictions aimed at the latter. This reads as a direct response to disputes that flared after the rise of Chinese models like DeepSeek and Kimi, when several US labs suggested rival models had been trained by distilling outputs from their own closed systems without authorization. The letter's position: address misappropriation through targeted legal and commercial mechanisms, not blanket restrictions on a technique the entire field depends on.
The letter arrives without a specific legislative or regulatory proposal attached. It is a positioning document ahead of anticipated action on AI policy in Washington, calling on lawmakers to expand compute access for startups and researchers, fund shared training datasets and evaluation frameworks, and avoid what it calls premature restrictions on open models. This should be treated less as a settled policy outcome and more as an indicator of where major infrastructure and chip providers want the regulatory conversation to land. Players like Nvidia, IBM, and Dell have direct commercial reasons to want open-weight ecosystems to flourish since a wider range of deployable models sells more compute and services regardless of which lab produced the weights.
Procurement teams weighing open-weight versus closed-model deployments should factor in that the policy environment favoring one approach over the other remains unresolved, and any restrictions on distillation or open releases could shift the economics of self-hosted AI within a single legislative cycle. The coalition's breadth, spanning cloud hyperscalers, chip makers, security vendors, open-source foundations, and venture capital firms, signals that the open-weight argument has moved beyond ideological advocacy into a commercial imperative for a significant segment of the AI infrastructure stack.
Our Take
The letter represents a notable moment in AI governance because it aligns companies that rarely agree on strategy. Meta and Microsoft compete directly on consumer and enterprise AI products. Nvidia and IBM compete on hardware and hybrid cloud. CrowdStrike and Palantir operate in adjacent but distinct security markets. Their joint signature suggests the economic gravity of open-weight deployment has become strong enough to override competitive friction, at least on this policy question.
Editorially, the security argument deserves scrutiny. The open-source security analogy held for operating systems and libraries where vulnerabilities are discrete, patchable, and observable. AI model behavior is probabilistic, high-dimensional, and often emergent; the claim that many eyes make vulnerabilities shallow has not been empirically validated for foundation models at frontier scale. The letter acknowledges the guardrail-removal problem but offers no mechanism for accountability when open weights are misused. Policymakers should press for concrete proposals on liability, traceability, and incident response before treating the security equivalence as settled.
The distillation defense is pragmatically sound. Blanket restrictions on a foundational ML technique would chill research and favor incumbents with the compute to train from scratch. However, the line between legitimate distillation and unauthorized capability transfer is porous in practice. Targeted legal mechanisms require clear definitions of what constitutes misappropriation, a threshold the industry has not yet agreed on. The letter's call for shared datasets and evaluation frameworks is the most actionable and least controversial ask; public investment in open benchmarks and compute access would benefit the entire ecosystem regardless of the open-weight outcome.
FAQ
Which companies and organizations signed the open letter?
The signatories include Meta, Microsoft, Nvidia, IBM, Dell Technologies, CrowdStrike, Palantir, ServiceNow, Hugging Face, Perplexity, Mistral, Andreessen Horowitz, Y Combinator, the Linux Foundation, and Mozilla, among others. The coalition spans cloud providers, chip makers, security vendors, open-source foundations, and venture capital firms.
What is the difference between open-weight and closed AI models?
Open-weight models publish their trained parameters for anyone to download, inspect, modify, and run on their own hardware. Closed models, such as those from OpenAI or Anthropic, restrict access to API-only interfaces where the underlying weights never leave the vendor's infrastructure.
How does the letter address security concerns about open models?
The letter concedes that released weights are beyond the developer's control and that guardrails can be removed. However, it argues defenders need access to capable models to detect AI-equipped threats, and that closed models create single points of failure. It draws a parallel to the open-source security argument but does not cite specific AI vulnerability data.
What is the letter's position on model distillation?
The letter distinguishes legitimate distillation, a standard technique for evaluation, validation, and capability transfer between models, from unlawful extraction of value from closed models. It urges targeted legal and commercial mechanisms against misappropriation rather than blanket restrictions on distillation itself.
What specific policy actions does the letter request?
The letter calls on lawmakers to expand compute access for startups and researchers, fund shared training datasets and evaluation frameworks, and avoid premature restrictions on open models. It does not attach a specific legislative proposal.
