The federal government's voluntary AI review framework takes effect today, the same week OpenAI's EU compliance statement leaves out the copyright chapter. Plus Microsoft's new agentic security system, Google DeepMind's whole-body robot model, a critical Rails security patch, and a wildcard ghost lineage found in human DNA.

Ashley Capoot, CNBC · Jul 31

Federal agencies' 60-day deadline for voluntary frontier-model review under Trump's AI executive order arrives August 1.

Tech Times · Jul 31

OpenAI's EU compliance statement omits the GPAI Code's copyright and training-data transparency chapter.

Lucas Ropek, TechCrunch · Jul 27

Microsoft debuts MAI-Cyber-1-Flash and Project Perception, agentic red/blue/green teams for security workflows.

Bloomberg · Jul 30

Google DeepMind's new model gives humanoid robots whole-body control and multi-robot task coordination.

Ruby on Rails · Jul 29

Security patch fixes a critical Active Storage flaw allowing arbitrary file read and possible remote code execution.

Sean Goedecke · Jul 29

Argues methodical, slow thinking under uncertainty distinguishes strong engineers from merely smart ones.

Simon Willison · Jul 31

Podcast appearance discussing Kimi K3's open-weight competitiveness and the week's AI security incidents.

GitHub Changelog · Jul 30

Rounds up July's Copilot updates, including generally available vision support and redesigned agent sessions.

Latent Space · Jul 30

Argues AI agents are driving renewed interest in ontologies and semantic-web-style structured data.