🔢少数派•Stalecollected in 73m
Fable 5 and Mythos 5 models pulled due to US ban
💡Understand how US export controls are forcing the immediate removal of high-performance AI models from the market.
⚡ 30-Second TL;DR
What Changed
Fable 5 and Mythos 5 models are no longer accessible.
Why It Matters
This development forces developers relying on these models to seek alternatives and underscores the growing geopolitical risks in AI infrastructure.
What To Do Next
Review your model dependencies and evaluate alternative open-source or locally hosted models to mitigate supply chain risks.
Who should care:Developers & AI Engineers
Key Points
- •Fable 5 and Mythos 5 models are no longer accessible.
- •The removal is directly attributed to U.S. Department of Commerce export restrictions.
- •This indicates a significant shift in compliance requirements for AI model deployment.
🧠 Deep Insight
Web-grounded analysis with 18 cited sources.
🔑 Enhanced Key Takeaways
- •The Fable 5 and Mythos 5 models, developed by Anthropic, were taken offline globally on June 12, 2026, merely three days after their public launch on June 9, 2026.
- •The U.S. Department of Commerce's export control directive cited national security concerns, specifically mentioning reports of a 'jailbreak' method for Fable 5 and suspicions of a China-linked group having accessed Mythos 5.
- •This incident represents the first time the U.S. government has issued an export control directive targeting an AI model itself, rather than just related hardware like semiconductor chips.
- •Anthropic had a prior dispute with the Trump Administration, having refused a Pentagon request to develop autonomous weapons systems, which resulted in the company being blacklisted by the Pentagon.
- •The directive mandated blocking access for any foreign national, including Anthropic's own non-citizen employees, leading the company to disable the models for all customers worldwide to ensure compliance.
📊 Competitor Analysis▸ Show
| Feature/Model | Anthropic Fable 5 / Mythos 5 | OpenAI GPT-5.5 (implied) | Anthropic Claude Opus 4.8 |
|---|---|---|---|
| Developer | Anthropic | OpenAI (implied) | Anthropic |
| Key Capabilities | Autonomous coding, complex reasoning, scientific research, vision, agentic hacking (Mythos 5) | Vulnerability discovery (implied) | General AI capabilities, less powerful than Mythos-class models |
| Benchmark Highlights | SWE-Bench Pro: 80.3% coding; FrontierCode: #1; CursorBench: 72.9% (outperformed GPT-5.5) | Outperformed by Fable 5 on CursorBench; capable of discovering vulnerabilities | Used as a fallback for sensitive Fable 5 queries |
| Pricing (per million tokens) | Input: $10, Output: $50 | Not specified in search results | Not specified in search results |
| Availability | Pulled globally due to US export ban | Generally available (implied) | Generally available |
🛠️ Technical Deep Dive
- Fable 5 and Mythos 5 are based on the same underlying 'Claude 5 base model,' sharing identical raw capabilities, parameter counts, and training.
- The primary distinction lies in behavioral constraints and safety safeguards: Fable 5 is the public version with strict inference-time controls, while Mythos 5 has these safeguards lifted for vetted partners.
- Fable 5's safeguards route sensitive requests (e.g., related to cybersecurity, biology, chemistry) to a less capable model, Claude Opus 4.8, in less than 5% of sessions.
- Mythos 5 possesses advanced cybersecurity capabilities, including the ability to discover and exploit software vulnerabilities and perform agentic hacking (reconnaissance, discovery, lateral movement).
- Both models are cloud-only, meaning there are no open weights available for download or local deployment.
- They are designed for long-running, autonomous tasks, demonstrating exceptional performance in software engineering, knowledge work, vision, and scientific research.
- Anthropic requires 30-day data retention for all Mythos-class model traffic on platforms like Amazon Bedrock to monitor and detect patterns of misuse.
🔮 Future ImplicationsAI analysis grounded in cited sources
Increased scrutiny and regulation of frontier AI models.
The unprecedented nature of this ban on an AI model itself, driven by national security concerns and 'jailbreak' fears, sets a precedent for stricter government oversight on powerful AI deployments.
Acceleration of AI sovereignty efforts outside the US.
The immediate global shutdown of US-developed models for all foreign nationals, including allies, will likely push other nations and companies to develop and rely on their own AI models to avoid similar disruptions.
Shift towards 'hybrid AI' architectures for businesses.
The abrupt removal of cloud-only frontier models will encourage businesses to adopt strategies that combine powerful cloud models for complex tasks with local, open-source models for everyday operations, reducing single-vendor dependency.
⏳ Timeline
2022-10
U.S. enacted comprehensive export controls on advanced semiconductor technologies.
2025-01
Biden administration introduced the AI Diffusion Rule, requiring licenses for advanced computing integrated circuits.
2025-05
Trump administration rescinded the AI Diffusion Rule.
2025-07-23
Trump administration issued Executive Order 14320, launching the U.S. AI Exports Program.
2026-05-31
U.S. Department of Commerce (BIS) issued guidance clarifying license requirements for advanced computing items, particularly for entities linked to China.
2026-06-09
Anthropic launched its Claude Fable 5 and Mythos 5 AI models.
2026-06-12
U.S. Department of Commerce issued an export control directive, leading Anthropic to disable Fable 5 and Mythos 5 globally.
📎 Sources (18)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
📰
Weekly AI Recap
Read this week's curated digest of top AI events →
👉Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: 少数派 ↗