Fable 5 and Mythos 5 Remain Offline — Anthropic Pushes Back Against Government Order
Four days after the US Department of Commerce issued an unprecedented export control directive, Claude Fable 5 and Mythos 5 remain completely unavailable to all users worldwide. The models launched on June 9 and were pulled just three days later on June 12, marking the first time a US government agency has forced a commercial AI model offline.
The order prohibits access by any foreign national — whether inside or outside the United States — including foreign-national Anthropic employees. Because Anthropic cannot filter users by nationality in real time, the company shut both models down globally to ensure compliance.
The trigger appears to be a jailbreak discovered by AI researcher Pliny the Liberator, who published a multi-agent attack strategy that circumvented Fable 5's safety classifier. The full 120,000-character system prompt has since been published on GitHub.
Anthropic issued a sharp response: "We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people. If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers."
All other Claude models, including Opus 4.8, remain fully operational. However, a Snyk analysis found that 16% of enterprises have no continuity plan for losing access to a key AI provider — a gap now painfully exposed.
The Trillion-Dollar IPO Race: Anthropic and OpenAI Both File with SEC
In what may become the most consequential capital markets event in tech history, both Anthropic and OpenAI have now filed confidential S-1 registration statements with the SEC within a single week of each other.
Anthropic filed first on June 1, fresh off a $65 billion Series H round that valued the company at $965 billion. OpenAI followed on June 8-9, with Goldman Sachs and Morgan Stanley managing its offering and a target valuation near $1 trillion.
The timing creates an extraordinary dynamic: two direct competitors, each seeking roughly the same valuation, racing to public markets simultaneously. Combined with SpaceX's Nasdaq debut on June 12 at $135/share (the largest IPO in recorded history), 2026 is shaping up as a watershed year for tech public offerings.
Anthropic president Daniela Amodei has publicly stated that compute costs are forcing the company toward an IPO, while OpenAI noted they may wait because "there are things we want to do that are likely easier as a private company."
Google's Gemini 3.5 Pro Nears Launch with 2M Context and Deep Think
Google's next flagship model, Gemini 3.5 Pro, is in its final stretch toward general availability, expected before June 30. Unveiled at Google I/O on May 19, the model features a 2-million-token context window and a new "Deep Think" reasoning mode — Google's answer to the extended reasoning capabilities pioneered by OpenAI's o1 series.
Key specifications:
Context: 2 million tokens — enough to hold entire codebases or book-length documents in working memory
Reasoning: Deep Think mode for complex multi-step problems
Pricing: ~$15/million input tokens, ~$60/million output tokens
Deep Think access: Gated to the $250/month Ultra subscription tier
Prediction markets currently put June 30 as the most likely GA date at 51% implied probability. CEO Sundar Pichai's I/O comment to "wait another month" reportedly drew audible groans from the developer audience.
Japan Deploys Claude Mythos Across Government and Banking
In a significant international AI deployment, Japan's government and its three megabanks — MUFG, Mizuho, and Sumitomo Mitsui Banking — have been granted access to Claude Mythos for financial cybersecurity applications.
Japan's Finance Minister Katsuki Katayama announced a public-private joint task force to assess Mythos's potential risks to the financial system, while Trend Micro confirmed it will use Mythos for cyber defense services. Hitachi has also joined Project Glasswing and will deploy Mythos Preview at its Cyber Center of Excellence for infrastructure vulnerability detection.
Anthropic simultaneously extended Mythos partnerships to approximately 150 new organizations across 15+ countries — though the Fable 5 shutdown raises questions about deployment reliability for international partners.
Microsoft's Surface Laptop Ultra: Running 120B-Parameter Models Locally
Announced at Computex on June 1, the Surface Laptop Ultra is Microsoft's first Windows PC powered by Nvidia's RTX Spark chip — a system-on-chip delivering 1 petaflop of AI compute with 128GB of unified memory.
The hardware can run AI models of up to 120 billion parameters entirely on-device, marking a significant milestone for local AI inference. Key specs include 20 CPU cores, 6,144 GPU cores, and a 15-inch mini-LED display hitting 2,000 nits peak HDR brightness.
The device targets developers and power users who need to run large models without cloud dependency — particularly relevant given events like the Fable 5 shutdown demonstrating the fragility of cloud-only AI access. Pricing and exact release date are expected later this year.