AI photo 'restoration' tools are fabricating history by inventing details that never existed, and experts warn that no workflow refinement can justify altering the visual identity of real people — true conservation means preserving what exists, not generating fiction.
Google launches Gemini 2.0 Flash Live via its Live API, giving developers the power to build real-time voice and vision AI agents with support for over 90 languages, reduced latency, and improved noise filtering across industries ranging from elder care companions to interactive gaming.
CapCut rolls out Dreamina Seedance 2.0 to paid users across seven markets, delivering AI-powered video generation with realistic motion, physics, and audio syncing for clips up to 15 seconds, backed by strict safety measures including real-face restrictions, invisible watermarking, and C2PA Content Credentials.
WhatsApp launches a major update introducing AI-powered writing assistance, in-chat photo editing via Meta AI, large file cleanup tools, cross-platform chat history transfer, and dual account support for iOS users, all while promising user chats remain private.
Charlotte robotics startup Lucid Bots raises $20M in Series B funding, bringing total investment to $34M, as its window-cleaning drones and robots surge from 100 to nearly 1,000 units sold while expanding into painting, waterproofing, and sealing markets.
Disney cancels its $1 billion OpenAI licensing deal after Sora's shutdown, though no money ever changed hands, and both companies are still exploring alternative partnership opportunities even as Hollywood's AI concerns shift toward rival platforms like SeeDance 2.0.
OpenAI shuts down its Sora AI video app just months after launch as Disney pulls a $1 billion investment deal, handing Google a major advantage in the rapidly evolving AI video generation market.
A custom C/Metal inference engine called Flash-MoE is now running a massive 397-billion parameter AI model on a standard MacBook Pro with 48GB RAM, streaming 209GB directly from SSD at 4.4 tokens per second — with 58 documented experiments revealing that Apple Silicon's unified memory architecture defies conventional optimization wisdom.
DoorDash is paying delivery workers to film their surroundings and record their voices for AI training through a new feature called Tasks, but security experts warn the biometric and spatial data collected can never be deleted once embedded in AI systems, raising serious privacy alarms for participants and bystanders alike.
Niantic Spatial is turning a decade of Pokémon Go player photos into a 30-billion-image map that now powers real-world robot navigation, with Coco Robotics already deploying the technology across 1,000 delivery robots in major cities worldwide.