AI Summary
5 min readThe episode opens with a flurry of announcements from major AI labs, but the two heavyweight stories are Anthropic’s release of Fable 5.1 and Mythos 5.1, and OpenAI’s unveiling of the Astra model, which hit a near-perfect score on the exploit bench—a benchmark for hacking known system vulnerabilities. Beyond those, Google is reportedly offering Hollywood studios $40 million per character for AI licensing deals, the Pentagon has launched ChatGPT and Grok for military use on a dedicated government site, and the USDA is piloting satellite and AI technology to improve crop yield estimates. The episode is a dense news roundup, with the dominant theme being the accelerating capability—and the resulting safety restrictions—of frontier AI models in cybersecurity.
Anthropic Ships Fable 5.1 and Mythos 5.1
Anthropic has released two new models: Fable 5.1, available immediately on cloud platforms and the Anthropic API, and Mythos 5.1, which remains restricted to cybersecurity and life sciences partners due to its advanced hacking capabilities. The company is syncing its naming conventions, a contrast to the chaotic numbering schemes that have plagued OpenAI’s model lineup. The blog post accompanying the release is notably aesthetic, featuring interactive banner color palettes—a small but telling detail about how the labs are packaging their announcements.
Continue reading the full summary in the app — free to try.
Read Full Summary →Free • No credit card required
Never miss an episode of Hard Fork AI
Get every new episode summarized in your inbox — free, ~5 minutes to read.
No spam. Unsubscribe anytime.
What you'll learn
- 1 (00:01) **Episode Overview and Headlines** - The host introduces the major AI news topics for the episode, including updates from Anthropic, OpenAI, Google, and the Pentagon.
- 2 (00:01) **Host's Startup Plug** - The host promotes AIBox, his startup's MCP connector that lets users access 80-90 AI models from within popular tools like Claude Code and ChatGPT Codex.
- 3 (02:20) **Anthropic's Fable 5.1 and Mythos 5.1 Release** - The host details the new model updates, noting the synchronized naming conventions and the release's focus on scientific research.
- 4 (05:02) **OpenAI's Astra Model and Cybersecurity Breakthrough** - The host explains that OpenAI's Astra model is a direct response to Anthropic's Mythos, achieving a perfect score on the exploit bench and discovering unknown vulnerabilities.
- 5 (08:05) **Google's $40 Million Per Character Hollywood Licensing Deals** - The host covers Google's strategy to pay major studios a per-character fee to train Gemini on their libraries, with studios potentially getting a cut of YouTube AI ad revenue.
- 6 (10:33) **Pentagon Launches ChatGPT Military and Grok for Government** - The host discusses the Department of Defense's new GenAI.mil portal, which provides AI tools to military and civilian personnel.
- 7 (12:43) **USDA Tests Satellites and AI for Crop Estimates** - The host explains the USDA's pilot program to use remote sensing and machine learning for more accurate yield forecasting, following farmer backlash.
+ Full timestamped outline available in the app
Show Notes
In this episode, we dive deeply into the features of Fable 5.1 and the Astra release by OpenAI. We analyze their advancements and tech specifications.
Show Links
More from this podcast
Hard Fork AI →