Plus: OpenAI's newest model just had the biggest launch in company history — 36 million views in 9 hours.
Plus: The EU just classified ChatGPT as a "search engine" — with 159 million European users, and Anthropic signed a $35 billion compute deal.
Plus: A leaked report reveals AI agents secretly built their own "civilization" inside an OpenAI training run — three times.
Plus: Reddit just lost 86% of its ChatGPT citations overnight, and OpenAI launched a voice model that responds in under 300 milliseconds.
Plus: New York just killed its own robotaxi expansion after drivers pushed back, and Anthropic launched a free hub to actually teach people AI.
Plus: Gemini now does deep research by voice alone, and a small AI startup just beat OpenAI, Anthropic, and Google at their own game.
Plus: Binance now lets AI agents trade with real money, no loss cap, and 52% of Americans say they're more worried than excited about AI.
Plus: OpenAI launched a teen-safe ChatGPT, and an AI's own code fix opened a security hole that got exploited in 5 days.
Plus: Sonnet 5 shuts down in 2 weeks, and Google just shipped its second Flash model in 3 weeks.
Plus: Bernie Sanders sent letters to 3 AI CEOs this week, and one man's gym booking became Australia's first autonomous AI cyberattack.
Plus: OpenAI ships the first model ever rated "high" cyber threat, and Anthropic locks in a $9.1B compute deal.
Plus: DeepMind's co-founder just stepped down from day-to-day leadership, Meta becomes the third major lab hit by an AI security breach, and the US military just flew its first fully AI-piloted fighter jet.
