Google DeepMind's AI Autonomy, Anthropic IPO Voting Control & the New AI Safety Group

Frontier labs are simultaneously consolidating founder power pre-IPO, self-regulating on safety, and racing on cost/compute — while DeepSeek scales revenue on high margins from selling access to its models.

• Recursive self-improvement, gradual: Google DeepMind's Koray Kavukcuoglu says agents now autonomously run training-loop experiments during multi-month model development, but always under human supervision — not end-to-end self-training.
• SAFA regulatory body advancing: OpenAI, Google and Anthropic are pushing an industry-led safety standards body (Standards Authority for Frontier AI) possibly launching by year-end, with Condoleezza Rice and David Friedberg floated as chair candidates; critics call it regulatory capture.

Themes, sentiment and reasons are AI-written paraphrases, checked so they never quote an episode. They may contain errors. Not investment advice. Episodes remain the property of their creators.