The US Commerce Department authorized Anthropic to distribute Claude Mythos 5 to a group of roughly 100 companies and federal agencies on June 27, partially lifting the export-control block imposed on June 12 (CNBC, CNN Business). Commerce Secretary Howard Lutnick signed a formal letter stating that appropriate safeguards were in place following negotiations between Anthropic and government officials (TechCrunch). The authorized pool covers members of Anthropic’s Project Glasswing - a vetted security-research program spanning many Fortune 500 firms - and the letter waives standard export-license requirements for foreign-national employees at those institutions (Semafor, NBC News). The June 12 directive had suspended all access after officials assessed that Mythos 5’s cybersecurity capabilities could accelerate vulnerability exploitation by adversarial actors (Axios). Claude Fable 5 remains suspended under the original order, with no published timeline for broader re-release (The Hill).
OpenAI reached general availability for Codex Remote on all paid ChatGPT plans on June 27 - covering Plus, Pro, Business, Enterprise, and Education subscribers - giving users phone-based control of long-running coding sessions on connected Mac or Windows hosts (TechTimes). The GA build replaces the previous shared-secret connection model with one-to-one QR-code pairing that enforces account matching and any configured MFA or passkey steps before granting phone access. A new DigitalOcean Droplet Workspace plugin lets Codex provision a cloud VM, configure SSH, and register it as a remote workspace from inside the app.
OpenAI also completed the retirement of GPT-4.5 from ChatGPT on June 26, concluding a 30-day sunset announced May 28 (OpenAI Help Center). Existing conversations previously tied to GPT-4.5 now route to GPT-5.5, and custom GPTs were migrated automatically. The model remains available via the API without change, preserving backward compatibility for developers running it outside the chat interface.