AI News Roundup – September 19, 2025
What does the future hold for generative AI?
Experts at an MIT symposium argued that future AI breakthroughs will shift from LLMs to ‘world models’ that learn autonomously through sensory input and interaction
massachusetts institute of technology
Sep 19, 2025
How AI startups are fueling Google’s booming cloud business
Google Cloud’s recent growth is being fueled by securing contracts with 60% of the world’s generative AI startups, attracted by cloud credits and GPU access
techcrunch
Sep 18, 2025
Amazon launches AI agent to help sellers complete tasks and manage their businesses
Amazon has launched an advanced AI agent for its third-party sellers that can proactively manage inventory, pricing, and advertising campaigns
techcrunch
Sep 17, 2025
Americans want AI to stay out of their personal lives
A new study shows that while Americans are comfortable with AI for large-scale analysis, a vast majority wants it to stay out of personal domains like religion and dating
theverge
Sep 17, 2025
Meta Connect 2025: The Evolution of AI Glasses, Metaverse Momentum, & More
At Connect 2025, Meta showcased new AI glasses, including a model with a display, as the key hardware for delivering personal superintelligence
meta
Sep 17, 2025
‘I have to do it’: Why one of the world’s most brilliant AI scientists left the US for China
Top AI scientist Song-Chun Zhu left the US for China to pursue his specific AGI vision, influenced by the resources available and a hostile climate for Chinese academics in the US
theguardian
Sep 16, 2025
The AI Ethics Brief #173: Power, Policy, and Practice
The latest AI Ethics Brief highlights the growing ties between Silicon Valley and the military and the risk of the ‘Companionship-Alienation Irony’ in AI companions
the montreal ai ethics institute
Sep 16, 2025
10 policy gold standards for AI leadership
Google has released a 10-point policy blueprint to help governments, especially in emerging economies, build a successful AI transformation strategy
google
Sep 16, 2025
Google releases VaultGemma, its first privacy-preserving LLM
Google has released VaultGemma, its first open-weight, privacy-preserving LLM that uses differential privacy to prevent the memorization of sensitive data
arstechnica
Sep 15, 2025
How people are using ChatGPT
A large OpenAI study reveals that most people use ChatGPT for everyday practical tasks, with 30% of consumer usage related to professional productivity
openai
Sep 15, 2025
AI Talks 