IC Technology & Surveillance — 2026-06-28

US Commerce Department Clears Anthropic Mythos 5 AI Model for Limited Release After NSA Breach Scare

BLUFClearing Mythos 5 while holding Fable 5 gives Anthropic a narrow cybersecurity lifeline but leaves its commercial trajectory and IPO timing hostage to undisclosed government criteria.

Commerce Secretary Howard Lutnick wrote to Anthropic's chief compute officer Friday, authorizing Mythos 5's release to "certain trusted partners" after the government suspended the model two weeks ago following an authorized NSA red-team evaluation on June 11 in which Mythos breached nearly all classified systems within hours, prompting export-control concerns that foreign nationals could bypass its security guardrails 123. Anthropic said restored access is limited to "a small group of cyber defenders and infrastructure providers," a partial resumption of the roughly 200-firm Project Glasswing program that had included Apple, Google, Cisco, Nvidia, and Microsoft 1. The action marks the first time the United States has applied export controls directly to an AI model rather than to the hardware or chips powering it; the Commerce letter makes no mention of Fable 5, which remains under separate restrictions 13. Bloomberg reported, citing people familiar with the talks who spoke anonymously, that the resolution followed direct meetings between Anthropic co-founder Tom Brown and Lutnick in recent days, with CEO Dario Amodei deliberately taking a hands-off role to reduce friction with the administration 1.

Analysis
The partial clearance formalizes a two-track deployment regime separating purpose-built cybersecurity tools from general-purpose frontier models, converting what began as an incident-response suspension into the first US export-control action targeting an AI model directly. Glasswing's restricted restart restores competitive footing for approved clients, but Fable 5's continued hold keeps Anthropic's broader commercial posture and IPO trajectory hostage to an undisclosed government timeline. Low confidence governs the assessment, reflecting the single-source evidentiary base and Commerce's omission of any stated criteria for restoring Fable 5. Brown's direct management of Lutnick negotiations while Amodei stepped back reads as founder-tier commercial leverage rather than technical compliance. The partial lift may reflect the competitive cost to US firms of sustained access restrictions more than genuine resolution of the government's underlying security concerns.
4 sources
  1. Anthropic Mythos 5 AI model cleared by US for wider use - Fortune
  2. Anthropic's Mythos 5 AI Model Cleared by US for Wider Use - Bloomberg
  3. Commerce Department greenlights partial return of Anthropic's Mythos - Axios
  4. Anthropic's Mythos 5 AI Model Cleared by US for Wider Use - Bloomberg

View in full brief →

UNCLASSIFIED // OPEN SOURCE