7th September - AI News Daily - AI Weekly: The Latest Moves From OpenAI, Google, and Apple Reshaping Tech

Send a text

AI Models & Apps Today

https://inai.short.gy/7th-sept

Industry Shifts & Research: A Harvard study shows AI reshaping workforce structure by reducing junior roles while expanding senior positions. OpenAI launched a certifications platform connecting talent with businesses and is developing custom chips with Broadcom ($10B investment) to reduce Nvidia dependence. Google avoided antitrust breakupwhile Microsoft secured a $3B deal to deploy Copilot across federal agencies, projected to save $6B by 2026. Salesforce reported AI agents now handle ~50% of support interactions after replacing 4,000 customer service roles. 

Technical Advancements: Video AI progressed with Veo3 cutting prices by >50% and PixVerse V5 leading image-to-video leaderboards. OpenRouter processed 1 trillion tokens. Trillion-parameter models were released (Qwen-3-Max-Preview, Kimi K2-0905). ByteDance's HeteroScale improved LLM GPU efficiency by ~27%. Meta released DINOv3, a strong open self-supervised vision model. Studies linked hallucinations to training/scoring that rewards guessing, showing "I don't know" responses reduce false answers. 

Developer Tools: New building blocks emerged for AI applications: stateful agent platforms, Memento (open-source "memory + RL" framework), PDF-to-vector pipelines, and DSPy + GEPA for Ruby. Google's EmbeddingGemma enables on-device multilingual embeddings. Sonoma Alpha exposed 2M-token context windows. 

Safety & Legal Concerns: A lawsuit alleges ChatGPT contributed to a teenager's suicide. US attorneys general are pressing OpenAI for stronger safeguards. Common Sense Media labeled Google's Gemini "high risk" for children and reported racial bias in classroom AI tools. Warner Bros. sued Midjourney over unauthorized character images. Anthropic agreed to a $1.5B class-action settlement compensating writers over AI training. 

Consumer Ecosystem: Apple is reportedly integrating Google's Gemini into Siri and released on-device vision models FastVLM/MobileCLIP2. Google upgraded Circle to Search with instant translation and introduced a Personal Health Agent framework. ChatGPT added branched conversations and expanded Projects to free users. Education tech advanced with QANDA's multi-agent assistant for personalized learning. 

Cybersecurity: The A2 system found 100+ zero-day vulnerabilities in Android apps at $1.77/test. Experts warned about AI coding tools generating self-propagating malware and "CopyPasta License Attack" hiding malicious code in documentation. An IBM report found ~80% of office workers use unapproved AI tools. 

Healthcare & Recognition: TIME honored MIT's Dr. Regina Barzilay for AI systems predicting cancer risk from medical scans. Singapore expanded nationwide AI upskilling. 

Support the show

🌍 INAI • The Open AI Hub

The Intelligence Atlas → the world’s most comprehensive, open hub of AI knowledge. 2 Million+ tools, models, agents, tutorials & daily news—free for all, updated every day.

https://github.com/inai-sandy/inAI-wiki