US restrictions on advanced model create opening for Chinese alternative

US restrictions on advanced model create opening for Chinese alternative

Sudden curbs on a leading American AI system highlight the appeal of fully open models that avoid unexpected global cutoffs for developers worldwide.

GP
Giulio Prisco
Jun 29, 2026
2 min read

The United States government recently directed a major artificial intelligence (AI) company to limit access to one of its most powerful systems. This action stemmed from concerns about vulnerabilities that could allow outsiders to bypass built-in safety rules, a process known as jailbreaking. As a result, Anthropic disabled its Claude Fable 5 model for users everywhere shortly after making a public version available in early June. The company had presented Fable 5 as an accessible form of its stronger Mythos technology, but the order forced a broad shutdown to meet export control requirements. Anthropic stated it disagreed with the move yet followed the rules while seeking ways to restore service for approved users.

Chinese AI developer Zhipu AI moved quickly in response, SCMP reports. The Beijing-based firm released its GLM-5.2 model just one day after the suspension took effect. Its founder described the American restrictions as deeply regrettable and pointed out that the new system is fully open. This means the underlying code is publicly available for inspection, modification, and use by anyone, without the same risk of sudden worldwide blocks.

Opportunity for open models from China

This timing drew attention from analysts who track the global competition in artificial intelligence. One observer noted that markets quickly recognized the gap left by the American cutoff and the chance it created for other providers. Developers who rely on advanced systems expressed worries about depending too heavily on United States companies, whose tools can face abrupt limits tied to national security rules. The episode has increased interest in alternatives that operate outside such constraints, even as American organizations continue to lead in the most capable frontier models, which represent the current peak of AI performance.

Zhipu has positioned GLM-5.2 as a practical option for those seeking stability and broad accessibility. In a later development, the United States government eased some of the earlier limits, CNBC reports, permitting restricted access to Fable for selected trusted organizations while broader availability remains under review. The situation shows how policy decisions in one country can shift user attention toward models from elsewhere that emphasize openness and independence from sudden external controls.

About the Writer

More from Mindplex

Keep reading

Three more ideas worth your time.

Browse News

Discussion

Join the discussion

Sign in to share a response with the community.

Type @ to mention someone Type / or use + to add a block Highlight text, then choose Link
Loading editor

Comments cannot be edited after posting because they become part of the reputation record. Give yours a quick review first.