Mythos 5 Cleared for 100+ US Companies; Economic Index Shows Claude Code Is the Most Automated AI Interface

TL;DR: The US government cleared Anthropic's Mythos 5 for 100+ companies and agencies after a two-week export-control standoff; Anthropic's June Economic Index shows AI usage mirrors real-world rhythms and Claude Code sessions are the most automated.

Mythos 5 unblocked for 100+ companies

The Trump administration resolved its standoff with Anthropic on June 26: Commerce Secretary Howard Lutnick sent a letter authorizing Mythos 5 access for "certain trusted partners" — more than 100 US companies and federal agencies, including Anthropic's own non-American employees (who were caught in the original ban).

Background: on June 13, the US government directed Anthropic to suspend all access by foreign nationals, citing national security export control authorities. That directive covered both Fable 5 and Mythos 5 and blocked non-US Anthropic employees from those models internally. Fable 5 approval is still pending.

Why it matters: Enterprise teams that had planned Mythos 5 deployments can now move — but the approved list is narrow (100+ named entities, not general public access). If your org was on hold, check directly with Anthropic.

Anthropic Economic Index: Cadences (June 2026)

Anthropic published the June 2026 Economic Index with hourly-resolution usage data and a 9,700-person survey. Key findings for builders:

• Claude Code is the most automated interface of any Claude product — agentic sessions run with far less back-and-forth than chat; work sessions overall skew more automated than personal ones
• Usage mirrors real-world rhythms: weekend personal use spikes to ~50% of conversations; recipe requests peak at 6pm (2.3× average); news queries at 7am; tax requests spike 8× around April 15
• 93% of conversations produce a concrete output — explanations (17%), documents (15%), guidance (11%)
• 35%+ of users expect AI to handle most of their work tasks within 12 months; optimism correlates with automation — those who delegate most express the most confidence in AI improving their pay and job security

Why it matters: The automation gradient data confirms what agent builders already sense — a Claude Code session is structurally different from a chat session. Budget your token and cost projections accordingly; agentic usage is consistently higher-volume than interactive use.

Sources: CNBC — Trump admin allows Mythos 5 release | TechCrunch | Semafor | Anthropic Economic Index — Cadences