On the evening of June 12, the U.S. Commerce Department issued a directive that required Anthropic to cut off every foreign national from its most capable AI models. The order, aimed at preventing potential security breaches, left the company with little choice: it pulled both Mythos—a preview model offered to roughly 200 organizations through the Glasswing program—and its sibling, Fable 5, from all users worldwide.
Anthropic’s internal systems could not segregate foreign users from domestic ones quickly enough. Rather than risk a partial compliance breach, the firm chose a blanket shutdown, taking the models offline for every customer, from Fortune‑500 firms to small startups. The move effectively ended three days of record‑setting benchmark performance by Fable 5, which had briefly topped every major AI test suite.
Only a few early testers kept limited access. Security companies Dragos and Cisco, both part of the original Glasswing cohort, continued to evaluate the models under special arrangements. Europe’s cybersecurity agency ENISA, which had recently joined the program, was pushed out by the same order, losing its preview access.
The incident underscores how AI’s dual nature—as a transformative commercial asset and a potential national‑security liability—has become a focal point for regulators. Anthropic’s co‑founder Jack Clark has publicly emphasized the company’s ongoing dialogue with the U.S. government about model safety, noting that the Commerce Department’s letter warned of criminal and civil penalties for non‑compliance.
Investors reacted swiftly. Shares tied to Anthropic’s pre‑IPO round slipped about 3.7% after the news broke, a modest dip given the firm’s rumored plans to pursue a valuation near $965 billion. The market’s sensitivity reflects broader concerns that the AI boom, once driven by abundant capital and flat‑rate pricing, is now confronting tighter financial and regulatory scrutiny.
Across the sector, bond issuers have raced to fund AI‑related projects. Man Group warned of bubble risks as bond sales surged to finance AI ventures, while utilities such as NextEra Energy have become leading hybrid‑bond issuers, financing massive data‑center expansions. The shift from free‑lunch, flat‑rate AI services to token‑based pricing models—exemplified by Google’s new AI Credits system—means enterprises are feeling the real cost of running large‑scale models.
Labor markets feel the ripple as well. Companies tout AI’s ability to boost productivity, yet the same technology threatens to displace workers in sectors ranging from legal services to manufacturing. European firms like Mistral, Siemens, and Schneider Electric are pushing AI onto factory floors to offset a shrinking workforce, while private‑equity groups pause sales of AI‑exposed assets out of caution.
Security concerns have risen in tandem with adoption. The trigger for Anthropic’s shutdown—an alleged vulnerability that could allow a jailbreak of Mythos—has sparked a scramble among firms to harden their data pipelines. FedHIVE’s CEO Michael Cardaci warned that the U.S. government wants American computing kept on American soil, a stance that could reshape global AI supply chains.
Anthropic’s experience illustrates a new reality: AI developers must balance rapid innovation with the risk of sudden regulatory intervention. As governments sharpen their focus on export controls, data sovereignty, and model safety, the industry may see more abrupt service disruptions, tighter licensing requirements, and a reevaluation of how AI products are priced and delivered.
Dieser Artikel wurde mit Unterstützung von KI verfasst.
News Factory APP - agentische News für besseres SEO & AEO.