Latest

AI Agents

NVIDIA, Meta and Microsoft back open AI without OpenAI, Google and xAI

NVIDIA brings together Meta, Microsoft and IBM to defend open AI in a manifesto without OpenAI, Google, Anthropic and xAI

NVIDIA, Meta and Microsoft back open AI without OpenAI, Google and xAI
NVIDIA, Meta and Microsoft back open AI without OpenAI, Google and xAI

NVIDIA CEO Jensen Huang has published his first message on X to publicly defend the role of open artificial intelligence models. He did so by sharing a letter titled Open Weights and American AI Leadership, signed by companies such as NVIDIA, Meta, Microsoft, IBM, Hugging Face, Mistral AI, Mozilla, Palantir and Perplexity, among others, but notable for the absences of OpenAI, Anthropic, Google DeepMind and xAI. The document argues that openness in AI models will be key to preserving the United States' technological leadership and to accelerating innovation, security and technological sovereignty.

AI will transform every industry, power every company and be built by every country, Huang writes. Open models strengthen security and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. The world needs both frontier closed models and frontier open models, he asserts.

The letter, dated today, July 24, 2026, draws a parallel between the free software movement of the 1980s and the current moment for artificial intelligence. According to the document, open source software made it possible to build much of the Internet, reduce development costs and create a shared technological base on which companies, universities and public administrations grew. The signatories believe that artificial intelligence is now going through a similar stage and that US leadership will not depend solely on having the most powerful model, but on creating an open ecosystem that allows AI to extend across the entire economy.

Defense of open models

The text argues that so-called open weights, models whose weights can be downloaded, inspected and run locally, make it easier for startups, universities, companies and administrations to develop solutions without depending exclusively on the major AI providers.

Spanish tourism versus AI agents: 12.6% of GDP at stake

The signatories argue that this openness fosters competition, reduces costs, accelerates innovation and prevents the market from concentrating in a small number of companies. The letter also states that openness improves security. While it acknowledges that open models can be misused, it argues that researchers and cybersecurity experts need access to advanced models to detect vulnerabilities, conduct defensive testing and develop protection systems against emerging threats.

Open AI versus closed AI

One of the manifesto's central messages is that security does not necessarily depend on keeping artificial intelligence models closed. The signatories argue that concentrating the most advanced capabilities in a small number of providers can create single points of failure, limit competition and hinder independent auditing. Against that model, they argue that open weights allow thousands of researchers, developers and cybersecurity experts to analyze systems, detect vulnerabilities and contribute to their improvement collaboratively.

The broad representation of the signatories, which includes chipmakers, AI model developers, software companies, open source organizations, investment funds and technology accelerators, reflects the growing support from part of the industry for an ecosystem in which state-of-the-art proprietary models coexist with open models. The goal, they argue, is to accelerate innovation, broaden access to artificial intelligence and prevent its development from being concentrated in the hands of a few companies.

Jensen Huang's publication of the manifesto comes at a time when AI has become a strategic asset for US industrial and technology policy. The document insists on the concept of technological sovereignty, arguing that governments, companies and institutions must be able to develop, deploy and control their own AI infrastructure without depending exclusively on a small group of providers.

Signatories:

  1. American Innovators Network
  2. Andreessen Horowitz (a16z)
  3. Arcee
  4. Arena
  5. Black Forest Labs
  6. Box
  7. CrowdStrike
  8. Dell Technologies
  9. Emergence
  10. Hugging Face
  11. IBM
  12. The Linux Foundation
  13. Mariana Minerals
  14. Meta
  15. Microsoft
  16. Mistral AI
  17. Mozilla
  18. NVIDIA
  19. Palantir
  20. Perplexity
  21. Reflection
  22. Replit
  23. ServiceNow
  24. Telnyx
  25. Y Combinator

Companies not among the signatories

  • OpenAI
  • Anthropic
  • Google DeepMind
  • xAI
  • Amazon
  • Apple
  • Oracle
Catalina Elidafamo

Author

Catalina Elidafamo

Journalist dedicated to covering the impact of innovation on markets and society, exploring everything from the potential of tokenization to the ethical dilemmas of AI.

The latest
From Huang to Botín, from Zara to Byron: corporate responsibility in global capitalism

Keep reading