Zuckerberg called Muse ready despite the risks, New York Times reports
THE SO WHAT
Meta greenlighting Muse over internal safety concerns under competitive pressure from a 14-person startup is the new AI release pattern—speed to market is being weighed directly against risk. If you're building agents, your real governance test is what you do when a smaller rival ships first.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIEU’s cyber agency and research centre are testing Chinese AI models
EU security and science arms quietly benchmarking open-weight Chinese models means model provenance is now a formal risk variable, not just a performance question. If you deploy open models in Europe, assume regulators will soon have an opinion on which families are acceptable for which data classes.
Applied AIMicrosoft Teams will soon be able to spot third-party deepfakes on your morning meeting
Deepfake detection moving into Teams is a sign that identity assurance is becoming a first-class feature of collaboration suites. CISOs should treat this as a baseline control and start mapping where else synthetic media verification needs to live across customer support, sales, and exec comms.
Applied AIGiorgia Meloni files to trademark her voice against AI deepfakes
A sitting prime minister moving to trademark her voice is an early template for personal IP defense against deepfakes. Public figures and brands should expect voice and likeness registries to become part of their risk stack—and start cataloging what needs protection before the legal norms harden.
Applied AIGoogle moves free Gemini to Flash-Lite and launches Nano Banana 2.1
Pushing free Gemini users down to Flash-Lite while rolling out a new Nano Banana 2.1 image model is Google drawing a sharper paywall between casual use and serious capability. If your product or workflow leans on "free" Gemini quality, expect degradation and either budget for paid tiers or diversify providers.