Studies question reliability of AI safety benchmarks

Stanford GSB fireside chat covers the World Development Report 2026, billed as the first comprehensive assessment of what AI means for developing economies. It arrives as development progress has dipped to its weakest pace in 75 years.

Speaking with Cleo Abram, the Google DeepMind chief argues AGI development should follow the scientific method — carefully, precisely and rigorously — with leading scientists involved.

Ibrahim moved from chief operating officer to a new C-suite title, chief AI readiness officer, at Google DeepMind, where she has worked since 2018. She discusses with CNBC's Julia Boorstin how to approach the ethical and existential questions surrounding AI.

Nano Banana 2.1 rolls out in the Gemini app, Search's AI Mode, Google Ads, AI Studio, Flow and Stitch. Google's own tests put it at 1,050 ELO for text-to-image preference vs 990 for Nano Banana 2 and 935 for Pro, with developer pricing starting at $0.0336 per standard 1K image.

Grok Imagine Video 1.5 Lite (SpaceXAI) entered Artificial Analysis's Text-to-Video Arena at #17 with a score of 985.4. MiniMax H3 Max leads the snapshot at 1117.1, ahead of MiniMax H3 (768p) at 1116.4 and Dreamina Seedance 2.5 at 1104.1.

Goldman Sachs' Kevin Sneader says the newest finance hires will oversee AI agents from the time they join rather than waiting years to become managers, raising questions about the fate of current middle managers.

Alphabet's AI drug discovery unit, spun out of Google DeepMind, is in early talks to raise new funds at a valuation of at least $40 billion, per people familiar with the effort.