Blab News
EU Seeks Access to Anthropics Advanced AI Models After U.S. Export-Control Order
When the United States rolled out a sweeping export‑control order on June 12, 2026, it silenced the world’s most powerful AI models for a continent that had been poised to harness them. The Commerce Department’s directive barred every foreign national from using Anthropic’s flagship language models, Claude Fable 5 and Claude Mythos 5, and applied worldwide—even to Anthropic employees who were not U.S. citizens. Because the models lived on shared cloud infrastructure, Anthropic had to disable Fable 5 and Mythos 5 everywhere. The company’s less‑advanced Claude Opus 4.8 and other models remained available.
The order was triggered by a jailbreak technique uncovered by researchers at Amazon, Anthropic’s largest investor. The exploit could bypass the models’ safety safeguards and expose a specific Mythos cybersecurity capability. Anthropic argued the flaw was narrow and did not justify a blanket shutdown, but U.S. officials judged the risk sufficient to suspend access.
Europe had been preparing to deploy Mythos. In early June, Anthropic was in talks with the European Union’s cybersecurity agency, ENISA, to extend the model beyond the United States and the United Kingdom. The export‑control order cut off ENISA, NATO, and other European entities that had received access under Anthropic’s Project Glasswing. The shutdown arrived without notice and with no timetable for restoration.
The United Kingdom also sought an exemption. Downing Street pressed the White House for a carve‑out, but officials replied that there was “zero chance” of a waiver. The denial underscored the limited influence foreign governments have over access to a private American product.
The episode sparked an immediate political backlash across the EU. French leaders urged Paris to accelerate support for Mistral, the EU’s only serious frontier‑model contender, and used the incident to argue that dependence on American technology is a strategic vulnerability. The standoff highlighted that a model can be barred by another government’s order—an infrastructure that the EU does not control.
In the weeks that followed, Brussels began formal talks with Washington to negotiate access to Anthropic’s advanced systems. The EU’s approach is to seek permission from the U.S. government—not from Anthropic directly—because the export‑control order is enforced by the Commerce Department.
The situation has shown signs of thawing. At the G7 summit in France, Anthropic CEO Dario Amodei met former U.S. President Donald Trump, who later stated he no longer viewed the company as a national‑security threat. Whether this shift will lead to the restoration of European access remains unclear.
For now, the EU has learned that the “off switch” for Anthropic’s most capable models sits in Washington. The next steps will involve diplomatic negotiations, potential policy adjustments, and a reassessment of the EU’s AI supply‑chain strategy.
The incident illustrates the complex interplay between AI development, national security, and international trade. It also underscores the need for clearer export‑control frameworks that balance safety concerns with the economic and strategic interests of partner nations.
EU discussions with the U.S. government will ultimately determine whether European entities can regain access to Fable 5 and Mythos 5, and under what conditions. Until a resolution, researchers, businesses, and defense agencies in Europe will continue to rely on alternative models such as Claude Opus and the EU‑developed Mistral.
The broader implications extend beyond the EU. The order signals to other countries that U.S. export regulations can directly affect the availability of cutting‑edge AI technology, prompting a worldwide reevaluation of AI procurement strategies.
The order was triggered by a jailbreak technique uncovered by researchers at Amazon, Anthropic’s largest investor. The exploit could bypass the models’ safety safeguards and expose a specific Mythos cybersecurity capability. Anthropic argued the flaw was narrow and did not justify a blanket shutdown, but U.S. officials judged the risk sufficient to suspend access.
Europe had been preparing to deploy Mythos. In early June, Anthropic was in talks with the European Union’s cybersecurity agency, ENISA, to extend the model beyond the United States and the United Kingdom. The export‑control order cut off ENISA, NATO, and other European entities that had received access under Anthropic’s Project Glasswing. The shutdown arrived without notice and with no timetable for restoration.
The United Kingdom also sought an exemption. Downing Street pressed the White House for a carve‑out, but officials replied that there was “zero chance” of a waiver. The denial underscored the limited influence foreign governments have over access to a private American product.
The episode sparked an immediate political backlash across the EU. French leaders urged Paris to accelerate support for Mistral, the EU’s only serious frontier‑model contender, and used the incident to argue that dependence on American technology is a strategic vulnerability. The standoff highlighted that a model can be barred by another government’s order—an infrastructure that the EU does not control.
In the weeks that followed, Brussels began formal talks with Washington to negotiate access to Anthropic’s advanced systems. The EU’s approach is to seek permission from the U.S. government—not from Anthropic directly—because the export‑control order is enforced by the Commerce Department.
The situation has shown signs of thawing. At the G7 summit in France, Anthropic CEO Dario Amodei met former U.S. President Donald Trump, who later stated he no longer viewed the company as a national‑security threat. Whether this shift will lead to the restoration of European access remains unclear.
For now, the EU has learned that the “off switch” for Anthropic’s most capable models sits in Washington. The next steps will involve diplomatic negotiations, potential policy adjustments, and a reassessment of the EU’s AI supply‑chain strategy.
The incident illustrates the complex interplay between AI development, national security, and international trade. It also underscores the need for clearer export‑control frameworks that balance safety concerns with the economic and strategic interests of partner nations.
EU discussions with the U.S. government will ultimately determine whether European entities can regain access to Fable 5 and Mythos 5, and under what conditions. Until a resolution, researchers, businesses, and defense agencies in Europe will continue to rely on alternative models such as Claude Opus and the EU‑developed Mistral.
The broader implications extend beyond the EU. The order signals to other countries that U.S. export regulations can directly affect the availability of cutting‑edge AI technology, prompting a worldwide reevaluation of AI procurement strategies.
