Anthropic’s Mythos 5 Returns — Partially The Verge / TechCrunch / CNBC After a two-week diplomatic standoff, the Trump administration has granted Anthropic permission to release Mythos 5 to roughly 100 US companies and federal agencies, including their non-American employees. The letter does not restore access to Fable 5, the public-facing Mythos-class model, which remains suspended for foreign users following a Commerce Department export control order triggered by a reported jailbreak attempt.

🔐 Policy & Export Controls

Trump Admin Releases Anthropic Mythos to 100+ US Companies and Agencies — TechCrunch AI

The US government has authorized a select group of companies and federal agencies to access Mythos 5, marking the first concrete step toward unwinding the emergency export control order. Non-American employees of those organizations are also covered. Access for everyone else — and all access to Fable 5 — remains frozen pending further negotiations.

Anthropic’s Mythos 5 Is Back — The Verge AI
The Verge obtained the government letter confirming the partial restoration; it emphasizes that the public-facing Fable 5 is explicitly excluded from this round of approvals.

White House AI Innovation and Security Executive Order — The White House
A broader executive order on AI innovation and national security was published this week, providing the policy backdrop for how the administration is framing export controls on frontier models.

🌏 Global AI Competition

Asian AI Startups Launch Mythos-Like Models as Anthropic’s Export Ban Drags On — TechCrunch AI

With Anthropic’s most powerful models restricted outside the US, a wave of Asian AI labs is moving fast to capture the gap. New frontier-capable models are launching with explicit pitches around their availability outside US jurisdiction — and analysts warn that US AI labs may struggle to recapture this market once users switch.

🤖 Frontier Models

Google Launches Gemini 2.5 Pro with Deep Think Reasoning Mode — Google

Gemini 2.5 Pro with Deep Think hit the leaderboard hard: 82.4% on GPQA Diamond, edging past Fable 5 (79.1%) and GPT-5.5 (76.3%) in scientific reasoning benchmarks. It is now the default in the Gemini app and AI Mode in Search, with API pricing at $1.50 / $9.00 per million tokens.

OpenAI Previews GPT-5.6 Sol, Terra, and Luna — OpenAI
OpenAI is previewing three new models under the GPT-5.6 family: Sol (frontier reasoning and long-horizon agentic work), Terra (GPT-5.5-competitive at 2× lower cost), and Luna (fastest and most affordable). Initial API access is rolling out to trusted partners and Codex users.

ChatGPT’s Global Market Share Falls Below 50% for First Time — Build Fast with AI
ChatGPT’s share of the global AI assistant market dropped to 46.4%, while Google Gemini rose to 27.7% and Claude reached 10.3%.

💼 Industry & Enterprise

Apple Vision Pro Exec Paul Meade Reportedly Leaving for OpenAI’s Hardware Team — TechCrunch AI

Paul Meade, the Apple VP responsible for Vision Pro, is joining OpenAI’s hardware division. The move adds a significant AR/VR product executive to OpenAI’s growing devices push and represents one of Apple’s higher-profile exits to a competitor this year.

Why Is Apple Asking Me to Pay More for Big Tech’s AI Obsession? — The Verge AI
Tim Cook called Apple’s pricing “unsustainable” and blamed AI infrastructure costs after the company raised prices on the MacBook Pro ($300), iPad Air ($150), and HomePod Mini ($30). The Verge pushes back on whether consumers should be subsidizing compute buildouts.

OpenAI Advertising Now Core Business as ChatGPT Tops 900M Weekly Users — MarketingProfs
With roughly one-fifth of ChatGPT queries expressing commercial intent, OpenAI has formalized advertising as a core revenue pillar alongside subscriptions and API.

🛠️ Developer Tools

Using Local Coding Agents — Ahead of AI

Sebastian Raschka’s deep dive into running open-weight models as local coding agents — covering tool selection, performance tradeoffs, and when a self-hosted setup can genuinely replace a Claude Code or Codex subscription. Practical and benchmark-grounded.

Claude Code 2.1.195 Changelog — Claude Code
Adds CLAUDE_CODE_DISABLE_MOUSE_CLICKS for fullscreen mode without mouse interference; fixes hook matchers with hyphenated identifiers (e.g., code-reviewer, mcp__brave-search) which were substring-matching instead of exact-matching; fixes voice dictation on macOS in long-running sessions.

📖 Society & Culture

Margaret Atwood Says the Problem with AI Is ‘Garbage In, Garbage Out’ — The Verge AI
At the Babel Literary and Cultural Festival in Porto, Atwood said she’s tried AI tools but remains skeptical — framing the core problem as the quality of training data rather than the technology itself.

The Fittest Founder in the Room Got Cancer. Here’s How He Used AI to Fight Back. — TechCrunch AI
Connor Christou fed his blood results, scan data, wearable output, and journal entries into Claude to help navigate his cancer treatment. A personal account of AI as a health research and decision-support tool.

Generated by claude-sonnet-4-6 on 2026-06-27T10:00:00Z