Anthropic's '5' era opens as Commerce lifts the Fable 5 / Mythos 5 export ban

breakthroughDeveloperLegalEnterprise

Wednesday, July 1, 2026

Confidence

High

Evidence

Commerce Secretary letter + company statements + Anthropic benchmark data + multi-outlet reporting

The White House reversed its 18-day export ban on Anthropic's two most powerful models the same night Anthropic launched Sonnet 5 — a cheaper mid-tier model that lands in the shadow of the returning flagships.

The details

The US government lifted export controls on Anthropic's most advanced models Tuesday evening; Anthropic said it "received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5" and would begin restoring access the next day. In a letter, Commerce Secretary Howard Lutnick said the tools no longer require an export license, and the firm agreed to work with the administration on safety rules and to report suspected malicious activity. CNN
The original ban was triggered when Amazon found a jailbreak around Fable's guardrails; Commerce had required suspending all use by foreign nationals — including Anthropic's own non-citizen employees — forcing the company to disable customer access to both models. Washington Post
On the same news cycle, Anthropic launched Sonnet 5 at $2 per million input tokens and $10 per million output through August 31, then rising to $3/$15 — cheaper than Opus 4.8 — and made it the default model for free and Pro plans. On one benchmark it scored 63.2% on agentic coding, versus Opus 4.8's 69.2% and Sonnet 4.6's 58.1%. TechCrunch
Sonnet 5's cybersecurity benchmarks came in worse than Sonnet 4.6, with Anthropic saying in its system card it "did not deliberately train" the model on cybersecurity tasks — an awkward note against the backdrop of the Mythos and Fable entanglement. Anthropic

What it means

The restoration is the fact, not the suspension. Any enterprise that paused Anthropic workloads just learned a single Commerce letter can turn a flagship model off worldwide, and that recovery rides on a private safety-cooperation deal customers never see. Sonnet 5's cut-rate window is a migration play — run a paired eval before the August 31 price step — but the real signal is that the government proved it controls the off switch on the operating layer everyone builds on. Write the export-shutdown clause the last contract omitted.

Counterview

Anthropic is a sui generis case — a weapons-capable model with a named jailbreak from a partner. Frontier labs without Mythos-class cybersecurity claims may never see this playbook applied.

The US government lifted export controls on Anthropic's most advanced models Tuesday evening; Anthropic said it "received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5" and would begin restoring access the next day. In a letter, Commerce Secretary Howard Lutnick said the tools no longer require an export license, and the firm agreed to work with the administration on safety rules and to report suspected malicious activity. CNN
The original ban was triggered when Amazon found a jailbreak around Fable's guardrails; Commerce had required suspending all use by foreign nationals — including Anthropic's own non-citizen employees — forcing the company to disable customer access to both models. Washington Post
On the same news cycle, Anthropic launched Sonnet 5 at $2 per million input tokens and $10 per million output through August 31, then rising to $3/$15 — cheaper than Opus 4.8 — and made it the default model for free and Pro plans. On one benchmark it scored 63.2% on agentic coding, versus Opus 4.8's 69.2% and Sonnet 4.6's 58.1%. TechCrunch
Sonnet 5's cybersecurity benchmarks came in worse than Sonnet 4.6, with Anthropic saying in its system card it "did not deliberately train" the model on cybersecurity tasks — an awkward note against the backdrop of the Mythos and Fable entanglement. Anthropic

Sources

Anthropic's '5' era opens as Commerce lifts the Fable 5 / Mythos 5 export ban

Turn a Payer Policy-Update Bulletin into Revenue-Cycle Action Items

OpenAI GPT-5.6 (Sol, Terra, Luna preview)