Recent actions by the United States government to restrict access to advanced artificial intelligence systems from Anthropic and OpenAI have triggered a growing shift toward open source AI models, particularly those developed in China. The moves have surprised the technology industry, which had become accustomed to AI companies releasing increasingly powerful models with little direct government intervention. The developments have also revived the long standing debate over closed and open approaches to artificial intelligence, raising concerns among businesses and developers about reliability, access, and control over the technologies they depend on.
Most of the world’s best known AI systems, including ChatGPT from OpenAI and Claude from Anthropic, operate as closed models. Their underlying code and training data remain inaccessible to the public, and users interact with the technology through applications, websites, or subscription services. The companies behind these models maintain control over access and can restrict availability when necessary. Open source or open weight models function differently because developers release the core files that power the systems, allowing anyone to download, modify, and run them independently. Once these models are released, they cannot easily be withdrawn or restricted by governments or companies. In early June, the Trump administration directed Anthropic to block access to its most advanced models, Mythos 5 and Fable 5, for non American users. Faced with the complexity of implementing user screening mechanisms, Anthropic reportedly took the models offline entirely. Shortly afterward, OpenAI agreed to a process in which the government would approve customers seeking access to its newest model, GPT 5.6.
The restrictions have highlighted concerns about depending on a single AI provider. Oren Michels, co founder and chief executive of Barndoor AI, said that businesses relying entirely on one frontier model face significant risks when access is suddenly interrupted. Haitham Mengad, co founder of AI powered music startup Stems Labs, said the removal of Anthropic’s Fable model was a significant moment that changed how he viewed open source alternatives. At the same time, open models have been attracting greater attention because the cost of using proprietary AI systems continues to increase. Around the period when restrictions on advanced AI models emerged, Chinese company Zhipu AI, also known as Z.ai, released GLM 5.2, an open model that reportedly performed close to leading systems from Anthropic and OpenAI across several benchmarks. Industry analysts noted that the ability to download, fine tune, and run the model on an enterprise’s own infrastructure could increase competitive pressure on companies developing proprietary models.
Usage trends also suggest growing momentum behind open alternatives. Data from OpenRouter, a platform that routes requests across different AI systems, showed that the combined market share of Google, Anthropic, and OpenAI declined from 55 percent to 33 percent between January and June. Chinese open model DeepSeek currently leads usage on the platform by a considerable margin. Industry observers note that organizations increasingly prefer flexibility and no longer want to depend entirely on a single AI provider. Among Western companies, French startup Mistral remains one of the few major advocates of open source AI, while Meta has reduced its emphasis on the approach despite previously supporting it. Concerns about Chinese models being potential security risks have also softened in some circles, with supporters arguing that once an open model is downloaded and operated on private infrastructure, its developer has no access to user data or control over how it is used. However, some experts believe that as open models become more capable, governments around the world may seek to impose restrictions on them as well, extending the debate over AI access and regulation beyond the United States.
Follow the SPIN IDG WhatsApp Channel for updates across the Smart Pakistan Insights Network covering all of Pakistan’s technology ecosystem.