Commerce Lifts Export Controls — Fable 5 and Mythos 5 Return July 1
The saga is over. On June 30, Anthropic announced that the US Department of Commerce has lifted the export controls on both Claude Fable 5 and Mythos 5, and access to both models officially resumes today, July 1. That closes the 18-day frontier blackout that began June 12, when the models were pulled after a jailbreak demo bypassed Mythos’s cyber classifier.
Monday’s narrow carve-out — Mythos back for ~100 vetted critical-infrastructure orgs only — has now widened to a full restoration for both models. The frontier reopens globally, not just through a government gate. The episode leaves a marker, though: for the first time, the state proved it can flip a frontier model on and off at will.
Anthropic Launches Claude Science for Researchers and Pharma
Anthropic released Claude Science, a research platform that puts a central AI assistant on top of 60+ scientific databases and computation tools in a single workspace — preconfigured for areas like genomics, proteomics, single-cell analysis, and cheminformatics. The assistant can spin up sub-agents for specialized tasks and is in beta for all Claude Pro, Max, Team, and Enterprise subscribers.
The play is explicit: automate the grunt work of research and accelerate drug discovery. Northeastern researchers say it could meaningfully speed up early-stage pharma work, and CEO Dario Amodei has long framed compressed scientific timelines as the whole point of the company. This is that thesis shipping as a product.
Claude Apps Gateway Comes to Amazon Bedrock and Google Cloud
Anthropic introduced the Claude apps gateway for Amazon Bedrock and Google Cloud — a self-hosted control plane for Claude Code that adds corporate SSO, centrally enforced policy, role-based access, per-user cost tracking, and spend caps. Teams get a single place to govern how their developers use Claude Code across cloud providers.
This is the natural next beat after last week’s per-agent cost attribution: the tooling is shifting from “let devs run agents” to “let the org run agents safely.” SSO and spend caps are the boring, essential features that turn a pilot into a rollout finance and security will actually sign off on.
Enterprise-Managed MCP Connectors Land, Starting with Okta
Admins can now provision MCP connectors for a whole organization through their identity provider, starting with Okta. Authorize a connector once, and users inherit access through the IdP groups and roles they already have — it’s just there the first time someone opens Claude. This is the first implementation of the Enterprise-Managed Authorization extension to the Model Context Protocol.
Asana, Atlassian, Canva, Figma, Granola, Linear, and Supabase support enterprise-managed auth at launch, with Slack coming soon. It’s in beta today for Team and Enterprise plans. Zero-touch connector setup is exactly the friction remover that gets MCP adopted beyond the early-adopter crowd.
Sonnet 5 Is Now the Claude Code Default; API Housekeeping Lands
Claude Sonnet 5 is now the default model in Claude Code, with a native 1M-token context window and promotional pricing of $2/$10 per Mtok through August 31 (update to v2.1.197). A new streaming idle watchdog is on by default for all providers, aborting and retrying when a stream produces no events for five minutes.
On the API, you’re no longer billed when a request returns stop_reason: "refusal" without generating output. And fast mode for Claude Opus 4.7 is deprecated, with removal on July 24 — migrate to fast mode on Opus 4.8 before then.
The Partner Ecosystem Lines Up Behind Managed Auth and Governed AI
The enterprise-managed auth launch arrives with a real bench: Asana, Atlassian, Canva, Figma, Granola, Linear, and Supabase already support it, a signal that the MCP connector ecosystem is consolidating around a shared enterprise standard rather than a scatter of one-off integrations. When the tools your team already pays for show up as governed, IdP-managed connectors, adoption stops being an IT project.
It fits the broader “governed AI” theme driving Claude’s enterprise momentum — the same demand pulling Claude into Snowflake Cortex AI and behind Anthropic’s $100M Claude Partner Network. The pattern is clear: the differentiator in mid-2026 isn’t raw capability, it’s control.
Claude Science Is the Bet That Justifies the Valuation
Strip away the drama of the export-control saga and the same-day Claude Science launch is the more consequential move. A near-trillion-dollar valuation can’t be underwritten by coding assistants and chat alone — it needs a story where AI creates categorically new value. “Automate research and compress drug-discovery timelines” is that story, and now it’s a shipping product with 60+ databases wired in, not a keynote promise.
The timing is almost poetic: the government spends 18 days demonstrating it can switch off Anthropic’s frontier models over cyber-misuse risk, and the day the controls lift, Anthropic ships the model into genomics and cheminformatics labs. That’s the tension of this IPO year in one week — the same capability that makes the state nervous is the capability being pointed at curing disease. Whether Claude Science delivers real scientific wins, not just faster literature reviews, is the question that actually decides if $965B was cheap or crazy.