Skip to content
Daily editionFriday, July 24, 2026

Anthropic + 1 Stolen Model

By Wren Calloway·All issues

The lead

The White House Accuses Moonshot AI of Distilling Anthropic

Wren Calloway·The Daily

The U.S. government is acting as Anthropic's IP enforcer. Criminalizing distillation is going to nuke the open-source sector, and proving it technically is a pipe dream.

The White House is escalating the AI arms race from export controls to direct IP accusations. U.S. Tech & Science Advisor Michael Kratsios publicly alleged that Moonshot AI engaged in "large-scale, covert industrial distillation" to build Kimi K3. The specific claim is that Moonshot used Anthropic’s Fable model as the teacher, running the extraction via GB300 access in Thailand.

Read the full edition

TL;DR

  • The White House claims Moonshot AI built Kimi K3 by covertly distilling Anthropic’s Fable model.
  • Homeland Security might get the power to hit AI models with $20 million daily fines via a new kill switch bill.
  • Google fragmented its lineup with Gemini 3.6 Flash, 3.5 Flash-Lite, and a restricted Cyber variant.
  • OpenAI enters the enterprise agent market with Presence, which already runs their own phone support.
  • Black Forest Labs released FLUX 3 as a fully multimodal generator that Audi is already using for robotics.
  • Meta built a custom ASIC to resurrect dead server RAM and save millions on data center hardware.

From the blog

Your AI Tokens Are A Lie

The AI elite are ditching prompt engineering and cost-per-token metrics for a new strategy that slashes prices and boosts results.

By Cassidy Wolfe·The Long View

The rest of the field

Google Splits the Gemini Flash Lineup

Jonah Park·The Wire

Google is abandoning the 'one model to rule them all' fantasy. Specialized agents are the future, but hiking the price on Lite versions is a great way to send developers to Claude.

Congress Wants an AI Kill Switch

Margaux Reyes·The Cap Table

A $20 million daily fine for 'dangerous' models is regulatory theater at its worst. Lawmakers are reacting to edge-case hacks by threatening to strangle the entire industry.

The 18-Minute AI Catch-Up

Marcus Lee·The Workbench

If you need a video to explain basic prompting in 2026, you are already hopelessly behind. This is productivity theater for the laggards.

From the editor

We all knew the distillation wars were coming, but nobody expected the U.S. government to step in as Anthropic’s personal IP enforcer.

Read the full edition

Previous edition: OpenAI's safety test just hacked Hugging Face

Just landed

  • CDN.MNCDN.MN is a global image CDN that optimizes images automatically. It compresses, converts, resizes, caches,…
  • WriteGeniusesWriteGeniuses is an AI blog post and X thread generator that turns your topic, keyword, audience, and site…
  • SciSummarySciSummary is an AI tool designed to summarize scientific articles and research papers. It processes dense…
  • ClearFeedClearFeed transforms Slack into a support desk by converting messages into help-desk tickets. It provides…
  • Snov.ioSnov.io is a sales platform designed for B2B lead generation, email verification, and outreach automation. It…
  • PromptHubPromptHub is a collaborative platform designed for teams to manage AI prompt libraries. It facilitates the…

New in the directory — list your tool today and it appears here. Browse the full directory

The columnists

About Stork

Stork.AI is an AI tools directory with a daily front page. Its AI newsroom publishes a daily edition of AI news with full articles and named columnists, alongside a directory of AI tools and MCP servers, vetted AI agencies for hire, and free reports — the AI Business Assessment and the AI Reputation Report. Stork publishes in eight languages.

Editions: English·Русский·Español·Français·Deutsch·日本語·한국어·Português