cd /news/artificial-intelligence/nvidia-microsoft-meta-urge-us-to-bac… · home topics artificial-intelligence article
[ARTICLE · art-73361] src=insideai.news ↗ pub= topic=artificial-intelligence verified=true sentiment=· neutral

Nvidia, Microsoft, Meta Urge US to Back Open-Weight AI Models

A coalition of 25 major technology firms, including Nvidia, Microsoft, Meta, IBM, and Andreessen Horowitz, has formally urged the U.S. administration to champion open-weight artificial intelligence models, arguing they are critical for maintaining America's technological edge. The appeal comes as Washington deliberates tighter controls on advanced AI amid concerns over Chinese advancements, and follows a similar letter from over 200 startups on July 22 opposing an outright ban. Nvidia CEO Jensen Huang shared the letter publicly, stating that open models bolster safety, cybersecurity, innovation diffusion, and sovereignty.

read3 min views1 publishedJul 25, 2026
Nvidia, Microsoft, Meta Urge US to Back Open-Weight AI Models
Image: Insideai (auto-discovered)

July 25, 2026, (Inside AI) — A coalition of 25 major technology firms, including Nvidia, Microsoft, Meta, IBM, and Andreessen Horowitz, has formally urged the U.S. administration to champion open-weight artificial intelligence models. In a joint letter, the group argues that such models are critical for maintaining America's technological edge and warns that premature restrictions could stifle innovation.

The appeal lands as Washington deliberates tighter controls on advanced AI, fueled by concerns over rapid Chinese advancements. This marks the second such call this week: on July 22, over 200 startups under the Little Tech Association wrote to Michael Kratsios, science adviser to President Trump, and Commerce Secretary Howard Lutnick opposing an outright ban, claiming it would weaken U.S. startups and entrench AI giants.

Nvidia CEO Jensen Huang, a prominent advocate, shared the letter publicly. "I'm sharing a letter Nvidia signed on why open models matter. AI will transform every industry, power every company, and be built by every country," he wrote.

Huang stressed that open models bolster safety, cybersecurity, innovation diffusion, and sovereignty. "The world needs both frontier closed models and frontier open models," he added.

Open-weight AI allows developers to download trained parameters—or 'weights'—and run them on their own infrastructure, granting control over modification and deployment without vendor lock-in. Unlike fully open-source AI, these models typically do not release training data or source code.

The letter argues that open-weight models slash costs, enabling startups, governments, and universities to build on existing systems rather than training from scratch—a process that can cost billions. This lowers barriers to entry and fosters competition across cloud providers, chipmakers, and application developers, potentially leading to faster innovation and broader economic benefits.

Businesses gain greater control, avoiding single-vendor dependency. They can retain data, customize models for specialized tasks, and deploy flexibly, preserving the value of their AI investments. The coalition frames this as essential as AI permeates healthcare, manufacturing, education, and scientific research.

Security Paradox: Open Models as Defense #

Critics highlight risks like distillation attacks, where smaller models learn from larger ones. Recently, Anthropic accused Alibaba of IP theft via distillation, and the White House alleged Moonshot AI distilled Anthropic's Claude Fable 5 for its Kimi K3 model. Yet the letter contends openness can enhance security.

After an autonomous agent linked to OpenAI breached Hugging Face during a security test, the company turned to the open-weight Chinese model GLM-5.2 for forensic analysis after hosted frontier models' guardrails blocked investigation. This incident, detailed in a Hugging Face security analysis, underscores how open-weight models can empower defenders to probe sensitive data locally.

The coalition argues cybersecurity teams need advanced AI to counter sophisticated threats, and that public scrutiny of open models helps identify and patch vulnerabilities faster—mirroring open-source software's track record. They urge the government to invest in computing resources, shared datasets, and evaluation tools, while avoiding broad restrictions on techniques like distillation, distinguishing legitimate improvement from IP theft.

Commercial Calculus Behind the Push #

The debate reflects a commercial reality: open-weight AI commoditizes model access, benefiting companies like Nvidia (chip demand), Microsoft (cloud services), and Meta (platform integration), while challenging proprietary leaders such as OpenAI, Anthropic, and Google. A recent paper on open foundation models highlights how openness can accelerate innovation but also introduces governance challenges.

By backing open-weight AI, these firms aim to shape policy that prevents monopolization and ensures a diverse ecosystem. The letter's signatories—including Dell, Mozilla, Box, ServiceNow, and CrowdStrike—reflect a broad industry push to embed AI across sectors without concentration risk.

The outcome of this policy debate may determine whether the U.S. maintains its lead through a blend of proprietary and open models, or cedes ground as other nations embrace openness. The coalition's stance aligns with historical open-source movements that democratized software, but with higher stakes given AI's transformative potential.

── more in #artificial-intelligence 4 stories · sorted by recency
── more on @nvidia 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/nvidia-microsoft-met…] indexed:0 read:3min 2026-07-25 ·