AI Export Controls: Fable 5 Ban, WorkOS Auth.md, and Future of Frontier AI

Introduction

The Fable 5 and Mythos 5 models were abruptly banned by the US government, raising concerns about the future of AI development and the accessibility of frontier AI. WorkOS's auth.md standard offers a solution to simplify secure agent registration for services, addressing a critical need in an increasingly regulated AI ecosystem.

Configuration Checklist

Element	Version / Link
Language / Runtime	Web application environment
Main library	WorkOS AuthKit
Required APIs	auth.md standard
Keys / credentials needed	Secure agent registration credentials (managed by WorkOS)

Step-by-Step Guide

Step 1 — Make Your App Agent-Ready

To enable services to register agents on behalf of users, implement the auth.md standard. This allows agents to interact with your application without needing to navigate traditional sign-up forms, streamlining automated processes.

# auth.md

Enable agents to register users without the sign-up form.
Auth.md provides secure agent registration that any app can implement.

## Make Your App Agent Ready
For services that want agents to register users on behalf of their customers.

Read the apps guide →

## Become an Identity Provider
For platforms whose agents act on behalf of users.

Read the provider guide →

## One-click enable with AuthKit
Get in touch to enable auth.md on your account.

Get early access →

Step 2 — Implement Agent Verified Flow

This flow allows an agent to register a full account by itself, providing agent-attested registration. This is useful for automated systems that need to create and manage accounts independently.

# auth.md

At its core, auth.md is a small Markdown file an application hosts at its domain — typically https://service.example.com/auth.md. The file describes how agents register with that service, which flows it supports, the scopes it exposes, and what to do once registered. Because Markdown is plain text, the same file works as documentation for human integrators and as a discoverable runtime artifact for agents. See how an auth.md is structured.

There are two primary flows for how an agent can register with an app:

## Agent verified flow
Agent-attested registration.

Step 3 — Implement User Claimed Flow

Alternatively, you can implement a user-claimed flow where the agent registers an account that is then claimed by a human user through a code-confirmation process. This balances automation with user control.

# auth.md

## User claimed flow
Code-confirmation registration.

Comparison Tables

Artificial Analysis Intelligence Index

Model	Intelligence Index	Reasoning Model	Open Weights / Proprietary
Claude Fable 5 (with fallback)	60	Yes	Proprietary
Claude Opus 4.5 (max)	56	Yes	Proprietary
GPT-5.5 (s-high)	55	Yes	Proprietary
GLM-5.2 (max)	51	Yes	Open Weights
Claude Sonnet 4.5 (flash)	50	Yes	Proprietary
Gemini 1.0 (flash)	50	Yes	Proprietary
Gemini 1.0 (pro preview)	47	Yes	Proprietary
Dwen7-7 Max	46	Yes	Proprietary
Minimax-M3	46	Yes	Proprietary
DeepSeek V4 Pro (Max)	44	Yes	Proprietary
Muse Spark	44	Yes	Proprietary
Kimi K2.6	43	Yes	Proprietary
Mixtral-8x22B-Pro	43	Yes	Proprietary
MiMo-V2.5-Pro	42	Yes	Proprietary
DeepSeek V4 Flash (max)	42	Yes	Proprietary
GPT-5.4 (s-high)	40	Yes	Proprietary
Gemini 1.0 (mini)	40	Yes	Proprietary

⚠️ Common Mistakes & Pitfalls

1. Unfulfilled Promises for Information: Anthropic initially promised to share more details about the Fable 5 ban within 24 hours, but this never happened. This lack of transparency can erode trust and leave users and developers in the dark. Fix: Establish clear communication channels and realistic timelines for updates, even if the news is unfavorable.
2. **Misinterpretation of