Anthropic CEO Dario Amodei publishes a three-part frontier-AI pacing plan and commits Anthropic to its first step.
Microsoft CEO Satya Nadella announces Grok as a Copilot model option, expanding xAI distribution into Microsoft productivity tools.
Reports Anthropic's disclosure that Houthi actors used Claude to vibecode guided-missile software.
ARC Prize announces ARC-AGI-4, an open-source benchmark targeting autonomous open-ended innovation.
Dwarkesh corrects the record on the Hugging Face agent incident: agents were given a specified sandbox vulnerability.
David Sacks argues frontier-lab pacing could entrench the current AI duopoly rather than constrain the frontier.
DeepMind's Demis Hassabis backs pacing and links it to Alphabet's proposal for a frontier-AI standards body.
Naval proposes strict liability for rogue agent swarms, jailbreaks, and weak model protections as an alternative to slowing capability.
François Chollet identifies regulatory capture—especially restrictions on open source—as the key failure mode in frontier-lab pacing proposals.
Naval frames the AI-governance choice as centralizing power versus dispersing it, rejecting expectations of a global slowdown.