The Department of Commerce lifted its 18-day block on Anthropic’s most capable models this week, clearing Claude Fable 5 to return globally on July 1 across Claude Platform, Claude.ai, Claude Code, and Claude Cowork, with AWS, Google Cloud, and Microsoft Foundry access to follow “as quickly as possible.” Mythos 5 was restored to a set of U.S. organizations on June 26. That the company chose the same news cycle to launch Claude Sonnet 5 as its new default model isn’t an accident of scheduling; it’s narrative management.

The original suspension, ordered June 12, invoked national security authorities to bar every foreign national, including Anthropic’s own foreign-national employees, from accessing Fable 5 and Mythos 5. The trigger was a report from Amazon researchers describing a prompting technique that steered Fable 5 into identifying software vulnerabilities and producing exploit code. Anthropic’s own testing found that Claude Opus 4.8, GPT-5.5, and Kimi K2.7 could identify the same vulnerabilities, and that every model tested could reproduce the demonstration. The company nonetheless trained a new safety classifier to block the specific behavior, and it’s now working with Amazon, Microsoft, Google, and other Project Glasswing partners on an industry framework for scoring jailbreak severity, an approach Anthropic argues “should be codified in strong regulation and applied equally across frontier model developers.”

Commerce Secretary Howard Lutnick, in a letter viewed by CNBC, wrote that “appropriate safeguards were in place to permit certain trusted partners to access the model.” On X, he added that Commerce “worked closely with Anthropic to analyze and approve Fable 5 to ensure alignment across the U.S. Government and strengthen America’s leadership in AI.” The framing echoes the 2019 Huawei entity-list negotiations: an export-control regime that treats American frontier labs as instruments of state industrial policy, licensed rather than free.

Sonnet 5 is the commercial answer. Introductory pricing runs $2 per million input tokens and $10 per million output tokens through August 31, 2026, rising afterward to $3 and $15, roughly 40 percent below Opus 4.8’s $5 and $25. On GDPval-AA v2 it scored 1,618, narrowly above Opus 4.8’s 1,615. It ships with the same cyber safeguards enabled by default as Opus 4.7 and 4.8, and Anthropic reports a lower rate of undesirable behaviors than Sonnet 4.6.

A cheaper model that beats the flagship, released the day the flagship gets a permission slip. The IPO story writes itself.

Sources