Episode Details

OpenAI is slowing parts of its frontier-model development to strengthen cyber safeguards, while separately outlining a privacy-preserving approach to safety monitoring for enterprise API users. We also look at a new Nature article on safety and security challenges for healthcare language models, plus brief updates from SpaceX, Anthropic, and the developer-tooling world.

Covered stories:
- OpenAI pauses parts of frontier reinforcement-learning training while hardening cyber safeguards
- OpenAI previews Private Safety Processing alongside Zero Data Retention for eligible API customers
- Nature article: safety and security of large language models in healthcare
- Cognition CEO denies a report that SpaceX sought to acquire the AI coding startup
- Report says Anthropic is using an unpublished internal model called Model 2
- smolvm sandbox testing for untrusted Python and JavaScript
- A case for LLM-assisted extensible software with sandboxed plug-ins

Sources:
- OpenAI, “Pacing model development in an era of cyber-critical capabilities”: https://openai.com/index/pacing-model-development-cyber-capabilities
- OpenAI, “Offering Zero Data Retention for frontier models”: https://openai.com/index/offering-zero-data-retention-for-frontier-models
- Nature, “Safety and security of large language models in healthcare”: https://www.nature.com/articles/s41586-026-10687-1
- TechCrunch, “Cognition CEO denies report that SpaceX tried to acquire the startup”: https://techcrunch.com/2026/08/19/cognition-ceo-denies-report-that-spacex-tried-to-acquire-the-startup/
- The Decoder, “Anthropic uses an unpublished AI model called Model 2 internally”: https://the-decoder.com/anthropic-uses-an-unpublished-ai-model-called-model-2-internally/
- Simon Willison on smolmachines and smolvm: https://simonwillison.net/2026/Aug/19/smolmachines-untrusted-sandbox/
- Simon Willison quoting Jeremy Morrell on extensible software: https://simonwillison.net/2026/Aug/19/jeremy-morrell/