Anthropic's restrictions are the best argument yet for sovereign AI, says Upstage CEO

When the US government ordered Anthropic to block foreign users from its most powerful models, the company took the simplest route: it switched them off worldwide at once instead of trying to build complex geo-restriction systems. For sovereign AI advocates, this was a ready-made argument in a long-running debate, not merely a news item.

A lesson that can't be ignored

Sung Kim, CEO of South Korean AI startup Upstage, seized on the situation publicly. He says Anthropic's response clearly demonstrates what proponents of local development have long warned about: foreign companies and governments cannot safely build critical infrastructure on American AI platforms, because they simply have no control over what happens next.

The sovereign AI debate has been going on for several years. Until now, though, it rested mostly on hypothetical scenarios: what happens if the US imposes restrictions? Now it has a real precedent: a global shutdown by one of the leading AI providers that hit users around the world without advance notice.

What happened to Anthropic's models

The US Department of Commerce expanded export controls on advanced AI technologies. The new rules covered Anthropic's highest-performing models, the ones used in enterprise applications, business process automation, and products with embedded AI.

Rather than build a complex system to verify users' locations, Anthropic chose an administratively simple fix: immediately disabling the affected models for all international customers. From a regulatory compliance standpoint, it was clean. For businesses that had already built products and workflows around these models, it was a serious blow.

Startups and mid-sized companies, which lack the resources to switch quickly to alternatives, were hit especially hard.

What foreign users got:

• Sudden loss of access to flagship models without warning
• No transition period and no migration guidance
• Outages in products and services that rely on the Anthropic API
• An urgent need to rethink their technology stack

Who came out ahead

Upstage is a South Korean company specializing in enterprise language models. Its flagship is the Solar series. Its key competitive advantage: customers can deploy the models locally, on their own servers, with no dependence on American infrastructure or US export laws.

Sung Kim was pushing the sovereign AI concept long before this incident, especially in Asian markets, where distrust of foreign tech platforms has traditionally run high. The Anthropic situation turned his arguments from theoretical into practical. Kim spoke out, calling the episode exactly the case the debate had long been missing: not a hypothetical, but one grounded in the real experience of thousands of companies that lost access to their tools in an instant.

For Korean, Japanese, European, and other companies that built infrastructure around American AI platforms, the episode was a stark reminder of the risks. For Upstage and other local players, it is the moment when potential customers are finally ready to listen.

What it means

US AI export policy is becoming a real operational risk for international business. Companies that built products on American models now have a concrete example: access can be revoked at any moment, not because of anything the company did, but because of geopolitical circumstances.

This will accelerate investment in local and regional AI development, especially in Asia and Europe. Diversifying AI infrastructure is shifting from an option to a standard risk-management practice.