Microsoft, tech companies throw weight behind spread of open-source AI
Microsoft, along with more than two dozen tech companies, are pressing policymakers to support open-source AI systems and code across society, arguing that it will be a safer approach than attempting to restrict access or relying on a handful of closed, proprietary models.
The open letter, posted Friday, draws parallels to the software industry of the 1980s, when large businesses worried that open-source software code would cut into their business. While industry lost that battle, the end result was a vibrant ecosystem that now underpins much of the modern internet, government IT and even commercial software products.
It also created a “shared foundation of knowledge” that has fed countless future software projects and innovations.
“The United States now faces a similar choice with artificial intelligence,” the companies wrote. “Our AI leadership will be judged not by one frontier AI model, but by whether the United States builds a strong, open ecosystem that diffuses into every sector.”
Expanding access and support to open-source AI comes with meaningful security risk. Cybersecurity experts warn that one of the biggest beneficiaries of broadly available AI tools are low-level criminals who until now lacked the technical expertise or resources to launch serious attacks.
Once a model is open weight, anyone can download it, customize it, strip it of any guardrails and use it for their own purposes. As open-source models have gotten better at creating deepfakes and other AI generated imagery, the danger of locally-customized CSAM and sexualized deepfakes could also grow.
But the letter argues that open-weight AI models are most beneficial to startups, universities, research labs and other small, ambitious organizations that can innovate and iterate the technology and make it more broadly useful to society.
“Open weights let every organization match the right model to the right job at the right cost, reserving frontier-scale capability for genuine frontier problems and running efficient specialized specialized models everywhere else,” The companies wrote. “That discipline is what will make AI economically sustainable as its use scales into the billions of everyday tasks.”
For cybersecurity specifically, the letter argues that defenders armed with open-source AI will outpace attackers better than any closed model approach.
“In a world where cybersecurity attackers use advanced AI, defenders need access to models with comparable capabilities so they can detect, simulate, and respond to emerging threats,” the companies wrote. “Open models broaden defensive capability, increase transparency, and allow vulnerabilities to be discovered and remediated across many teams.”
Other notable companies signing the letter include Meta, Palantir, Perplexity, Mistral, NVIDIA, Mozilla, The Linux Foundation, Hugging Face, Dell Technologies and IBM.
US policymakers continue to grapple with balancing unrestrained support for the domestic AI industry and providing oversight and regulation of harms that result from their use.
The Trump administration has cycled through several frameworks since coming into office, first a laissez-faire approach within no restrictions, then an executive order creating a voluntary testing regime for industry, then the imposition of export controls on Anthropic’s Fable model and reportedly pressuring OpenAI to delay the release of their models out of cybersecurity concerns.
The letter comes as the Trump administration has reportedly considered an executive order that would restrict American access and availability to Chinese-made open-source models.
But the White House and US companies are trying to thread a needle in recognizing the overall benefits of an open source approach while being wary of doing anything that could potentially benefit their Chinese rivals.
Earlier this month the White House announced the creation of its Gold Eagle AI cybersecurity clearinghouse that would help coordinate government, private sector and civil society work finding and closing AI-discovered vulnerabilities. A big part of that effort, a senior White House official said, is supporting providers and maintainers of open-source AI tools.