Trump’s Pulling the Plug on Us
By punishing Anthropic, the U.S. government could free Europe from its toxic addiction.
On Friday, June 13, Anthropic received an export control order to suspend access for non-U.S. citizens to Fable 5 and Mythos 5, the latest updates of its AI model. An export control law was invoked with immediate effect. The veto was so general that it even removed access for non-U.S. citizen Anthropic employees in the U.S. Given the technical impossibility of discriminating against users by country, the company disabled the two models, stating: “We believe this is a misunderstanding and are working to restore access as soon as possible.” A misunderstanding is a situation in which someone interprets something differently from how it was said, but the government’s letter did not give an exact explanation.
The order states that AI models are a technology that is subject to export controls and forces Anthropic to obtain licenses for any access by “foreign persons” in any part of the world, because they pose an “unacceptable risk” to national security. A typical move for Benjamin Netanyahu and Xi Jinping. But the order does not provide any criteria or technical tests to help solve the problem. For example, why Anthropic’s AI models are dangerous while OpenAI’s are not.
Alex Stamos, former chief security officer at Facebook and current chief information security officer at cybersecurity company SentinelOne, explained in a letter signed by hundreds of high-level specialists that Mythos is very good at finding vulnerabilities and exploiting them, but it is not the only [AI model] with those capabilities. He also explained that the domestic Fable model has such aggressive protections built into it that it is “the source of humor in the cyber community.” If the criteria were technical, then other models would have been blocked too. If it were specific, Anthropic could fix it right away. The ambiguity was deliberate and effective. The government is acting according to criteria known only to the government itself, and therefore these questions cannot be answered.
Many believe that Anthropic CEO Dario Amodei was asking for trouble, using fear as a marketing strategy in 2026. “One reaps what one sows,” as Yann LeCun said. Others claim that it is another attempt at extortion by the Department of Defense to force [AI companies] to abandon the restrictions on their models’ use, which continue to select and kill people on their own initiative, and to conduct mass surveillance of U.S. citizens. In any case, it is severe punishment for a local business leading in the most important race on the planet, especially after announcing its success in the stock market.
But also for the American industry, which has taken a reckless gamble with the promise of a new, automated economy and is running the existential risk of pushing the market toward Chinese and/or open models to avoid depending on a government prepared to lose the match in order to punish one player. And a warning for us: Which European company or administration can risk relying on Anthropic or any U.S.-owned platform, knowing that Donald Trump could boycott it at any moment?
Our governments, hospitals, universities, banks and businesses rely on the cloud and Amazon Web Services, Microsoft Azure and Google Cloud, as well as the following operating systems: Windows, Android and OSx. The EU Commission has just presented its tech sovereignty strategy to reduce its dependency on the cloud, AI, chips and software from the U.S. France is moving to Linux by decree-law. Let’s not be the last ones. Let’s not wait for the worst to happen.
On Friday, June 13, Anthropic received an export control order to suspend access for non-U.S. citizens to Fable 5 and Mythos 5, the latest updates of its AI model. An export control law was invoked with immediate effect. The veto was so general that it even removed access for non-U.S. citizen Anthropic employees in the U.S. Given the technical impossibility of discriminating against users by country, the company disabled the two models, stating: “We believe this is a misunderstanding and are working to restore access as soon as possible.” A misunderstanding is a situation in which someone interprets something differently from how it was said, but the government’s letter did not give an exact explanation.
The order states that AI models are a technology that is subject to export controls and forces Anthropic to obtain licenses for any access by “foreign persons” in any part of the world, because they pose an “unacceptable risk” to national security. A typical move for Benjamin Netanyahu and Xi Jinping. But the order does not provide any criteria or technical tests to help solve the problem. For example, why Anthropic’s AI models are dangerous while OpenAI’s are not.
Alex Stamos, former chief security officer at Facebook and current chief information security officer at cybersecurity company SentinelOne, explained in a letter signed by hundreds of high-level specialists that Mythos is very good at finding vulnerabilities and exploiting them, but it is not the only [AI model] with those capabilities. He also explained that the domestic Fable model has such aggressive protections built into it that it is “the source of humor in the cyber community.” If the criteria were technical, then other models would have been blocked too. If it were specific, Anthropic could fix it right away. The ambiguity was deliberate and effective. The government is acting according to criteria known only to the government itself, and therefore these questions cannot be answered.
Many believe that Anthropic CEO Dario Amodei was asking for trouble, using fear as a marketing strategy in 2026. “One reaps what one sows,” as Yann LeCun said. Others claim that it is another attempt at extortion by the Department of Defense to force [AI companies] to abandon the restrictions on their models’ use, which continue to select and kill people on their own initiative, and to conduct mass surveillance of U.S. citizens. In any case, it is severe punishment for a local business leading in the most important race on the planet, especially after announcing its success in the stock market.
But also for the American industry, which has taken a reckless gamble with the promise of a new, automated economy and is running the existential risk of pushing the market toward Chinese and/or open models to avoid depending on a government prepared to lose the match in order to punish one player. And a warning for us: Which European company or administration can risk relying on Anthropic or any U.S.-owned platform, knowing that Donald Trump could boycott it at any moment?
Our governments, hospitals, universities, banks and businesses rely on the cloud and Amazon Web Services, Microsoft Azure and Google Cloud, as well as the following operating systems: Windows, Android and OSx. The EU Commission has just presented its tech sovereignty strategy to reduce its dependency on the cloud, AI, chips and software from the U.S. France is moving to Linux by decree-law. Let’s not be the last ones. Let’s not wait for the worst to happen.
Previous article![]( "US–Iran: A Deal That Smells of Gunpowder")
