Is History Repeating Itself With American AI? – Inside Sources

There is a quote attributed to Mark Twain that goes, “History doesn’t repeat itself, but it often rhymes.” 

Having worked in the computing industry for more than 40 years, the debate over whether the United States should continue to focus almost exclusively on promoting closed-source artificial intelligence or embrace open-source models reminds me of a similar debate in the 1990s.

The struggle then centered on an aggressive campaign against Linux and the emergence of open-source operating systems and software. It took close to 20 years for companies wedded to a proprietary model to completely change their philosophy toward open source. Microsoft — once the chief antagonist — went on to embed a Linux kernel in Windows and become one of the globe’s largest corporate contributors to open-source software.

Today, American technology companies are confronting fast-moving global competitors from China that are releasing state-of-the-art open-source AI models. As the earlier open-source debate showed, no matter how much companies like Microsoft, or now Anthropic, want to sell expensive, monolithic, proprietary tech stacks, that approach alone will not be sufficient to maintain America’s lead in the global race for AI dominance.

Open-source AI models today serve as a powerful engine for domestic and global economic growth. Small businesses, startups and local innovators often lack the resources to license expensive proprietary models or build large-scale AI infrastructure independently. AI entrepreneurs can instead download, fine-tune and run open models on their own infrastructure, making them attractive to startups seeking lower costs and strong performance.

Closed corporate models, meanwhile, concentrate control and revenue inside a handful of firms, while open systems diffuse capability, invite oversight and mobilize the full depth of U.S. technical talent against both domestic challenges and Chinese competition. Closed models will continue to play an important role, but they cannot substitute for a strong American open-source ecosystem.

Prominent venture capitalist Marc Andreessen framed the AI industry’s competitive stakes bluntly: America must lead in open-source AI or risk a future in which “the entire world, including the U.S., runs on Chinese software.” He made that statement in May 2025, before the release of Beijing-based Moonshot AI’s Kimi K3. As Axios reported, it “has caught up to models that defined the U.S. frontier just weeks ago, at a substantially lower price.” In effect, Kimi’s release “reset the AI race overnight.”

Given the gap between American open-source AI models and their Chinese counterparts on key performance metrics, many American businesses and universities have turned to Chinese-developed models. Without a vibrant U.S. open-source ecosystem, American users will increasingly depend on foreign models — and on the technology stacks, standards and security risks that come with them.

As Misha Laskin, the CEO of Reflection AI, an American open-source model supplier, recently warned, Chinese open-source models can serve as “Trojan horses” for foreign technological ecosystems. They can potentially lock nations into proprietary hardware such as Huawei chips, which DeepSeek V4 was optimized to use, and there is the risk of “sleeper cell” deceptive behaviors embedded in these models. These concerns strengthen the case for trusted American alternatives and clear security, transparency and intellectual property standards for models used by U.S. businesses and institutions.

Unfortunately, the United States does not have the luxury of taking two decades to embrace open-source AI, as it did with open-source software. Some of the loudest warnings come from dominant closed-model companies whose revenues and market power depend on customers paying for access to proprietary systems. At today’s pace of development, we cannot allow such self-interest to delay the urgent work of jump-starting American alternatives capable of competing with China.

That’s why many of the largest technology companies, NVIDIA, Microsoft, Palantir, SpaceX, Perplexity, IBM to name a few, created the Open Secure AI Alliance to confront this risk. They urged the government to preserve open-weight models as a foundation for American leadership, instead of fearing them the way that technology companies did in the 1990s.

There needs to be a pro-export, pro-innovation strategy for American open-source AI today — one that harnesses U.S. strengths in computing power, talent and entrepreneurship. By championing open-source AI, America can reinforce its tradition of technological leadership, expand economic opportunity and ensure that the transformative benefits of artificial intelligence remain broadly accessible while reflecting the nation’s enduring values.