California Governor Gavin Newsom just vetoed SB 1047, blocking what would have become the strictest AI safety law in the US. The bill would have required models costing over $100 million to implement kill switches and testing protocols. Newsom argued the broad rules would stifle innovation, leaving frontier AI developers free from binding state restrictions.
Google just rolled out Gemini-powered contextual Smart Replies to Gmail on Android and iOS. Instead of generic one-liners, the AI analyzes the entire email thread to draft detailed responses complete with greetings and sign-offs. Why it matters: It turns quick replies into genuinely usable drafts, though it is currently limited to English and paid Gemini subscribers.
Google just added support for public YouTube URLs and audio files directly into NotebookLM. You can now analyze lecture transcripts, connect insights across recordings, and easily share generated Audio Overviews via a public link.
Google just officially named and released pre-trained weights for AlphaChip, its reinforcement learning system that treats chip floorplanning like a board game. The AI generates superhuman circuit layouts in hours instead of months, and has already designed Google's last three TPU generations as well as MediaTek chips. Why it matters: AI is no longer just running on advanced silicon—it is actively redesigning the hardware that powers it.
Microsoft just introduced a "correction" preview in Azure AI Studio that flags inaccuracies and rewrites flawed outputs before users ever see them. It works by grounding generated text against source documents, though Microsoft admits using models to police other models isn't foolproof.
Snap just expanded its partnership with Google Cloud to power Snapchat's My AI chatbot with multimodal Gemini models. Users can now send photos and videos to get instant help like translating street signs or picking healthy menu items, which Google says has already driven a 2.5x jump in U.S. engagement.
Google just rolled out updated versions of its Gemini 1.5 Pro and Flash models, boosting math reasoning by around 20% and cutting unnecessary refusals. The company also slashed 1.5 Pro input token costs by 64% and output tokens by 52%, while ramping up API rate limits. Why it matters: Developers can now run faster and sharper frontier models at substantially lower operating costs.
Google CEO Sundar Pichai just announced a $120M Global AI Opportunity Fund at the UN Summit of the Future. The company plans to partner with local nonprofits to bring AI education and training to communities worldwide, aiming to stop an AI divide before it widens.
ETH Zurich researchers just fine-tuned a YOLO vision model to crack Google's reCAPTCHAv2 image challenges with a 100% success rate. The model trained on around 14,000 labeled image pairs of road objects while mimicking human mouse movement and device fingerprinting. Google noted that most websites have already migrated to its invisible v3 system.
Google just rolled out major AI updates to Google Ads at DMEXCO, giving advertisers more creative control. Marketers can now supply up to five reference images to direct image generation, while Performance Max is gaining brand guidelines and campaign-level negative keywords by year's end.
Google just rolled out up to 10 different voice options for Gemini Live, allowing users to preview and swap tones anytime directly in the app. Why it matters: giving AI assistants a wider range of vocal personalities makes real-time voice conversations feel noticeably more natural and personalized.
Google just announced plans to flag AI-generated and edited images in Search, Google Lens, and Circle to Search later this year. The system relies on C2PA metadata to trace origins. It is a solid step against deepfakes, though missing support from tools like Flux remains an issue.
Google just brought Gemini straight into Chrome's address bar, letting users query the AI by typing "@gemini" followed by their prompt. While it puts generative AI just a keystroke away for millions, it currently lacks page context, meaning it can't analyze or summarize the tab you're actively reading.
Apple just rolled out its Clean Up tool in the latest iOS 18.1, iPadOS 18.1, and macOS 15.1 developer betas. Powered by Apple Intelligence, it lets you tap, circle, or brush away photobombers and clutter while automatically filling in backgrounds and clearing shadows. For now, it is limited to devices set to English and the U.S. region.
Google and Tel Aviv University researchers just unveiled GameNGen, a neural model that simulates the 1993 classic Doom in real time at over 20 frames per second using Stable Diffusion. In blind tests, human raters struggled to distinguish the generated video from the real game. It points to a future where games aren't programmed line by line, but hallucinated on the fly by neural networks.
Google just launched "Gems" for Gemini alongside its upgraded Imagen 3 image generation model. Subscribers can now build custom AI personas with plain-text instructions, while getting noticeably crisper, photorealistic visuals. Gems and human image generation are rolling out first to Advanced, Business, and Enterprise tiers.
Google just rolled out a "take notes for me" feature in Google Meet, using Gemini AI to capture key discussion points rather than generating a raw transcript. Summaries automatically save to a Google Doc that syncs with participants or calendar invites. For now, it is limited to English on computers for select Workspace add-on subscribers.
Google just rolled out three experimental models: an ultra-compact Gemini 1.5 Flash-8B, a significantly upgraded Flash, and a stronger Pro model that levels up coding and complex prompt reasoning. Both maintain 1M token context windows and are available to test for free in Google AI Studio and the Gemini API.
Google just appointed former Character.AI CEO Noam Shazeer as co-technical lead for Gemini, joining heavyweights Jeff Dean and Oriol Vinyals. Shazeer recently returned to Google following a massive $2.7 billion licensing and talent deal to turbocharge its fight against OpenAI.
Nearly 200 Google DeepMind employees just signed an internal letter demanding leadership terminate contracts with military organizations and cut off defense access to its AI. The staff cite Google's original 2014 acquisition pledge to never use DeepMind tech for warfare, underscoring growing friction with cloud initiatives like Project Nimbus. Google has yet to provide a meaningful response.
