Andrej Karpathy joins Anthropic to build Claude-powered pre-training team, Gemini 3.5 Flash API goes GA with aggressive pricing, Pentagon diversifies AI partners, and GitHub suffers security breach via VS Code extension.
Google unveils Gemini 3.5 and personal AI agent Gemini Spark, Andrej Karpathy joins Anthropic from OpenAI, and OpenAI adds C2PA watermarks to AI-generated images.
OpenAI ends its Microsoft exclusivity to bring models to AWS and Google Cloud, Anthropic overtakes OpenAI with $30B ARR, and Google rolls out proactive Gemini assistance.
Anthropic confirms testing Claude Mythos, Google launches Gemini 3.1 Flash Live, Microsoft pairs GPT with Claude to reduce hallucinations, and Apple plans to open Siri to rival AI services.
This week's AI developments: Google's Gemini AI now scans the dark web for threats, OpenAI builds a unified desktop 'superapp' to rival Anthropic, Canvas launches an AI teaching agent for higher education, and the Anthropic vs Pentagon legal battle intensifies.
This week's AI developments: OpenAI releases GPT-5.4 with native computer use, Google Maps gets a Gemini-powered Ask Maps feature, and Anthropic launches a code review tool for AI-generated code.