summree
Claude Fable JUST got BANNED...
Anthropic
Wes Roth

Claude Fable JUST got BANNED...

⏱ 22 min video · 3 min read13 Jun 2026Worth watching
TL;DR
The US White House directed Anthropic to suspend all access to Claude Opus 5 (codenamed Fable 5) and Claude Sonnet 5 (Mythos 5) for non-US nationals after Amazon researchers demonstrated a jailbreak exposing security vulnerabilities. Anthropic publicly disagreed with the decision while complying, and the creator also covers Claude 5's remarkable capabilities including autonomous coding, 3D game generation with self-directed debugging, and a 9.5-hour agentic software build.
Key points
1
The White House ordered Anthropic to suspend all Claude Opus 5 and Claude Sonnet 5 access to non-US nationals, effectively forcing Anthropic to pull access for everyone globally.
2
Amazon researchers triggered the government action by demonstrating a jailbreak that revealed known security vulnerabilities; Anthropic publicly stated the vulnerabilities were minor and that other public models can find them without any bypass.
3
Anthropic reversed its controversial 'invisible refusals' policy for frontier AI development queries, committing that all safeguard triggers will now be visible and fall back to Opus 4.8, just like cyber and bio safeguards.
4
Claude Code creator Boris Churnney called Claude 5 the biggest model step-up since Opus 4.5, praising its methodical debugging, judgment, and taste as emergent properties not engineered through prompting.
5
Ethan Mollick described a shift from 'wizard casting spells' to 'commissioning work,' and Claude 5 autonomously built a 19-page design document and executed a complex AI calibration tool called Concord over 9.5 hours from a single prompt.
Key arguments
If you are a non-US user, access to Claude Opus 5 and Claude Sonnet 5 is currently suspended globally until Anthropic resolves the government directive.
Anthropic has now made all safeguard triggers visible, meaning any refusal or capability restriction will be explicitly shown rather than silently degrading output quality.
Claude 5 can autonomously build production-level software tools from a single prompt over multi-hour sessions, which has immediate implications for researchers and developers needing expensive judgment-based data analysis pipelines.
Notable quotes

We've learned to grow intelligence in a petri dish. Only instead of a petri dish, it is a data center, but we grow it like we can grow bacterial colonies or fungus or gardens or whatever.

I don't even have any sort of benchmarks, any challenges I can give it that I don't think it would be able to solve.

Dario, on one hand, is asking for the governments to step in and regulate more. But he out of all the AI leaders seems to be butting heads the most with various branches of government.

Worth watching?
Worth watching the full video?
Watch if you want to see the actual Claude 5 gameplay demos and 3D generation in action — the key news and arguments are all captured here, but the visual proof of Claude 5 building shadowed 3D spaceship environments and self-debugging in real time is more striking to see directly.
Topics
AI & TechAnthropic

Explore more summaries on these topics →

Saved you some time? The creator still deserves a like.

Watch on YouTube →
More like this

Want this for your own channels?

Add the channels you follow. Every new video summarised and in your inbox the moment it drops. From £4/month.

Try it free