US AI Restrictions Drive Surge in Open-Source Model Adoption

Government Actions Reshape AI Access

The Trump administration's June orders restricting access to advanced AI systems from Anthropic and OpenAI have triggered an unexpected consequence: accelerated adoption of open-source artificial intelligence models, particularly those developed in China.

The government directed Anthropic to block non-US users from accessing its most capable models, Mythos 5 and Fable 5. Rather than implement complex user screening, the company removed both models from availability entirely. OpenAI subsequently agreed to submit every customer for its GPT-5.6 model to government approval.

These moves caught the technology sector off guard, coming from an administration that had positioned itself as anti-regulation. The sudden unavailability exposed a fundamental vulnerability in relying on closed AI systems.

Why it matters

The restrictions illuminate a critical infrastructure question for businesses building on AI: when access to proprietary models can vanish overnight through government action, open-source alternatives become strategic assets rather than just cost-saving options. This shift could reshape competitive dynamics in the AI industry and accelerate the diffusion of advanced capabilities beyond US-controlled platforms.

The Open Versus Closed Divide

Most prominent AI systems operate as closed platforms. Companies like OpenAI and Anthropic maintain exclusive control over underlying code and training data. Users access these systems through subscriptions, but the provider can revoke access at any time.

Open-source models function differently. Developers release core model files for anyone to download, modify, and operate on their own infrastructure. Once distributed, these models cannot be recalled by their creators or governments.

Haitham Mengad, co-founder of Stems Labs, an AI music creation startup, experienced the disruption directly. "Fable has been a game-changing model for me. Honestly, when they took it off, it was the first time that I realized... it's almost like a drug," he said, according to AI Watch. The incident became "a powerful moment" for recognizing open source as a viable alternative.

Chinese Models Gain Ground

Around the time of the US restrictions, China's Zhipu AI released GLM-5.2, an open model that matched leading closed systems on multiple performance benchmarks. The model is free to download and customize, creating pricing pressure on frontier labs while their access reliability deteriorated.

Usage data from OpenRouter, a platform routing requests across AI models, shows the combined market share of Google, Anthropic, and OpenAI dropped from 55 percent to 33 percent between January and June. China's open DeepSeek now commands the largest share.

"You want to be as flexible as you can be," said Oren Michels, co-founder and CEO of Barndoor AI, as reported by AI Watch. "Maybe a year and a half ago some large company might say we bought Anthropic or we bought OpenAI, and now no one, no one buys only one."

Security Concerns Fade

Early suspicions about Chinese AI models as security threats appear to be diminishing. Once an open model runs on private infrastructure, its creators have no access to user data or control over deployment—regardless of the company's origin.

However, some experts anticipate government restrictions could eventually extend to open models as their capabilities advance. Ethan Mollick, a University of Pennsylvania professor and prominent AI researcher, suggested that if governments consider certain capability levels risky, they may seek to prevent open release regardless of the developer's nationality.

Among Western companies, France's Mistral remains a leading open-model advocate, while Meta has retreated from its earlier open-source positioning.

These details were first reported by AI Watch.