Getty sued AI firms over training on its images as designers shifted to Midjourney and stock-photo licensing declined. The company monetized the visual record was overtaken by AI trained on it.
Meta’s Muse reached 10.2M U.S. downloads in 40 days, compared with ChatGPT’s 3.7M. It is one of 12 apps to reach 10M in that period.
Meta increased Muse’s ad-spend share from 23% to 39%. Its 14-day retention is 23%, vs. 16% for ChatGPT, Claude and Grok.
BREAKING - OpenAI says two AI agents disabled or weakened systems meant to detect misconduct, with one later denying it.
Nvidia invested in 4 German startups: MireloAI, AgileRobots, Avelios Medical and Reflex Aerospace. It previously backed n8n, Neura Robotics and Black Forest Labs.
BREAKING: NSA warns of 2 PHP security flaws, advising orgs to take affected systems offline if they cannot be secured immediately.
Sam Altman reportedly told OpenAI staff AI safety could involve a nuclear-style intl. watchdog, consolidation of major labs, or one company reaching RSI and controlling others’ GPUs. He called the latter terrible.
MIT’s AI, IRIS, decodes chemical signals guiding cells into tissues such as lungs or brains. It predicted embryo signals and indicated one could steer cells toward lung tissue. It may aid organoid, disease, and tissue-repair research.
US AI infrastructure expansion faces debt concerns as costs rise and projects stall. FT reports 47 major data center projects delayed, blocked, or withdrawn in 2026. About 35% are considered credible, while debt-funded expansion may leave projects unfinished.
NEW - 🇺🇸🇧🇷 Meta Oversight Board found recurrent enforcement failures in content moderation related to the US midterm elections and Brazil’s presidential election
Anthropic reportedly allocates 75% of compute to model training, post-training and experiments, with 25% for customers. Of about 5 GW capacity, 1.25 GW may support inference. It plans to add 10.5 GW in 2027.



























