By The Diplomat | Created at 2026-06-24 15:21:55 | Updated at 2026-06-25 18:18:49
1 day ago
On June 7, SK Telecom and Nvidia announced a partnership to build a gigawatt-scale AI cloud in South Korea. The deal positioned the country’s largest wireless carrier as a cornerstone of Seoul’s national AI strategy. Five days later, the White House ordered Anthropic to suspend all foreign access to its most advanced AI models. SK Telecom was also at the center of it.
The Trump administration’s June 12 export control directive forced Anthropic to disable Fable 5 and Mythos 5 for all users worldwide. The directive followed two separate incidents that eroded White House confidence in Anthropic. Days earlier, U.S. officials had ordered Anthropic to revoke access to Mythos for SK Telecom, which had joined Project Glasswing in early June. Officials cited the carrier’s alleged ties to China. SK Telecom denied the allegations, stating that the claims lack verified facts. Even after Anthropic complied shortly after, Amazon flagged a guardrail bypass, known as “jailbreak,” in Fable 5 to the White House, prompting the broader export control order.
Within the span of a single week, South Korea’s largest telecom secured a landmark AI infrastructure deal with Nvidia and then lost access to the foundational software it needed to make that infrastructure competitive.
For Anthropic, the irony runs deeper than bad timing.
The company has spent years building a public identity around the argument that Washington must use its regulatory power to constrain the global spread of advanced AI technology. In April 2025, Anthropic published an official position paper urging the Commerce Department to strengthen the AI Diffusion Rule, which was the Biden-era framework that imposed export controls on advanced AI chips. The paper argued that maintaining the United States’ compute advantage through export controls is essential for national security and economic prosperity. It called for tighter thresholds, stricter enforcement, and more funding for export control agencies.
Nvidia pushed back publicly, accusing Anthropic of telling tall tales and of trying to manipulate regulators rather than compete on merit. The exchange laid bare how corporate self-interest shapes export control advocacy in Washington. Anthropic relies on Nvidia hardware to train its model. Restricting Nvidia’s overseas chip sales artificially constrains demand, lowering chip prices for Anthropic’s domestic compute needs.
Federal records show Anthropic has spent nearly $5 million on federal lobbying since 2023, with the Commerce Department’s AI Diffusion Rule among its primary targets. But now the logic Anthropic helped build has turned against it. When the government applied that same national security logic to Anthropic’s own software exports, the company dispatched executives to Washington and contested the decision publicly.
The episode reflects a broader pattern in which U.S. export control decision are shaped less by coherent grand strategy than by the lobbying priorities of competing Silicon Valley firms. The result, as Chatham House pointed out in its critique of U.S. AI export controls, is that “Washington’s recent policy has been inconsistent and mercurial.”
For allied governments and firms, the inconsistency is not merely puzzling but costly. Multi-billion-dollar infrastructure decisions require regulatory predictability. When export controls shift in response to corporate feuds and unverified security concerns, the planning horizon for allied partners collapses.
South Korea’s position puts it squarely in the line of fire.
Beijing is preparing to spend $295 billion over the next five years on a coordinated AI data center network, with a mandate that at least 80 percent of supporting technology must come from domestic suppliers. The plan effectively closes China’s AI buildout to Korean firms. Samsung and SK Hynix supply much of the world’s AI memory, yet they cannot compete in a market engineered to exclude them.
That forced exclusion from China makes access to the U.S. ecosystem not a preference but a necessity. Korean firms must build within the U.S. AI supply chain or risk being squeezed out of both major markets simultaneously. That’s why the SK Telecom-Nvidia partnership is so important: It is a multi-year infrastructure bet on the reliability of the U.S. technology ecosystem.
SK Telecom’s subsequent exclusion from Mythos calls that reliability into question. SK Telecom’s China footprint is minimal by any commercial measure. It earned roughly $1.9 million in Chinese revenue in 2024 and has just seven employees in the country, according to its annual report. Its historical ties to China Unicom, through a joint venture stake that was sold in 2009, were treated by U.S. officials as grounds for suspicion. If that standard of scrutiny applies to one of South Korea’s most integrated companies, the threshold for allied access to frontier U.S. AI technology is effectively undefined.
Washington has long positioned export controls as a tool to protect allies and advance shared strategic interests. The Mythos incident suggests a different reality, one in which those controls can be triggered by unverified allegations and corporate complaints, with allied partners having to absorb the collateral damage.
South Korea cannot pivot back toward China’s AI ecosystem. Beijing’s domestic mandate has seen to that. For Korean firms building their AI future on U.S. technology, the Anthropic saga demonstrated that access to that technology can vanish overnight – not because of anything Seoul did, but because of corporate maneuvering by Silicon Valley.
