How Anthropic may have talked itself into an AI export ban

Anthropic’s repeated public warnings about the dangers of advanced artificial intelligence have become politically charged after Washington barred foreign nationals from accessing its newest models, Mythos and Fable, last week. An analysis by the Financial Times found that five in every 1,000 words used by Anthropic in official statements, social media posts and articles in 2026 related to risk, regulation or restrictions [1]. The equivalent figure for rival OpenAI and its chief Sam Altman was 0.6 words per 1,000, an eightfold difference [1]. The FT created lists of terms including “harmful,” “dangerous” and “misaligned” and calculated their frequency, supplementing the count with sentiment analysis [1]. Some technologists have blamed the export decision on the $965 billion AI group’s repeated warnings about AI’s risk to society, particularly in relation to Mythos [1]. Yann LeCun, Meta’s former chief AI scientist, wrote in a social media post that the ban showed Amodei’s “ridiculous fear-mongering” about AI had finally paid off [1]. “One reaps what one sows,” LeCun wrote [1]. The dispute has alarmed parts of Europe and Silicon Valley, where executives and officials fear the Trump administration may be willing to restrict non-US access to frontier models [1]. It is emerging as an early test of how the US intends to oversee increasingly powerful AI systems [1]. The broader landscape has already been reshaped by the rapid ascent of generative AI applications. Throughout the 2020s, tools such as ChatGPT and DALL-E became widespread, allowing users to instantly generate sophisticated texts, images and media [4]. The competitive dynamics intensified in early 2025 when DeepSeek, a chatbot developed by a Chinese company, surpassed ChatGPT as the most downloaded freeware app on the US iOS App Store [2]. DeepSeek’s success was described as “upending AI” and initiating “a global AI space race” [2]. Its compliance with Chinese government censorship policies and its data-collection practices also prompted regulatory scrutiny in multiple countries [2]. Meta itself has been a central actor in the open-model movement. Its Llama family of large language models, first released in February 2023, initially restricted model weights to researchers on a case-by-case basis under a non-commercial license [8]. Unauthorized copies of the first model were shared via BitTorrent, and subsequent versions were made accessible outside academia under licenses permitting some commercial use [8]. In April 2026, Meta Superintelligence Labs released Muse Spark as a replacement for Llama [8]. The US export restrictions arrive amid a decade marked by escalating military conflicts and a technological landscape in which multiple analysts increasingly refer to China as a superpower due to its influence in geopolitics, technology and manufacturing [4]. The G7 leaders, meeting in Évian-les-Bains, France, from 15 to 17 June 2026, issued joint statements on Ukraine, the Middle East, critical minerals and global economic imbalances, but the summit agenda did not publicly address the emerging AI export-control dispute [3].