Google DeepMind

Cyber attacks are becoming more sophisticated than ever. But what if AI could be our strongest defense? 🛡️ VP of Security John "Four" Flynn joins Hannah Fry on our podcast to explore how we're leveraging new and improved models to build a more secure digital world. Watch Part 1 of their 2 part conversation now. → https://goo.gle/46W659z

We're working with Commonwealth Fusion Systems, one of the world’s leading nuclear fusion companies, to speed up the development of clean, safe, limitless energy with AI. ⚛️ Fusion powers the sun and has the promise of abundant clean energy. However, one approach involving controlling the super-heated plasma needed for fusion is a monumental scientific challenge. Our collaboration focuses on accelerating the timeline to viable fusion energy in 3️⃣ key areas:  Fast, accurate simulation: we've released TORAX, an open-source plasma simulator allowing CFS to run millions of virtual experiments to test and refine plans for their powerful tokamak, SPARC.  Optimizing for maximum energy: using reinforcement learning and TORAX, we’re now rapidly identifying the most efficient ways for SPARC to generate more power than it consumes Real-time control: we’re developing AI agents that act as expert "pilots," learning to dynamically control the plasma in real time to manage immense heat loads and improve safety. This partnership combines our world-leading AI with CFS's state-of-the-art hardware to help build the foundations for a more sustainable energy future. Find out more → https://goo.gle/4n6Wph2

Veo is getting a major upgrade. 🚀 Here are some new ways you can create and edit videos with AI ⬇️ We’re rolling out Veo 3.1, our updated foundation model, which comes with richer realism, more natural audio and better image-to-video capabilities. To help you craft better stories, this model will power new tools like: 🖼️ Ingredients to video: Prompt with multiple reference images and Veo will create a final scene that looks as you envisioned - complete with audio. 💡 First and last frame: Create a smooth or epic scene that bridges your preferred start and end images ↔️ Scene extension: Veo can extend your clips to develop longer videos, even lasting for a minute or more, that continue the action from your original shot. These are just a handful of the updates rolling out today in our AI filmmaking tool Flow, and beyond. Find out how to get started →  https://goo.gle/4qdrwKD

Proud & excited to share that our Experience AI program is one of only four initiatives recognized by UNESCO for promoting the responsible use of AI in education! We created this program in collaboration with our partners at the Raspberry Pi Foundation to offer educators the lesson plans, classroom resources, and hands-on activities that help 11-14 year-old students develop a foundational understanding of AI & how it influences their daily lives. The program has reached more than 2 million students in 24 countries, and growing! We made a major bet on this program years ago because we know that giving young people access to the right skills & education is the best way to empower them. It’s been fantastic working with colleagues across the Raspberry Pi Foundation, Google, and Google DeepMind to ensure the content is as useful & relevant as possible. Congratulations to all our partners on this well-deserved recognition!  https://lnkd.in/eZaFzp-W

We’re proud to announce that Genie 3 has been named one of TIME's Best Inventions of 2025. Genie 3 is our groundbreaking world model capable of generating interactive, playable environments from text or image prompts. Find out more → https://goo.gle/3KGqiYa

Our new Gemini 2.5 Computer Use model can navigate browsers just like you do. 🌐 It builds on Gemini’s visual understanding and reasoning capabilities to power agents that can click, scroll and type for you online - setting a new standard on multiple benchmarks, with faster speed. We’ve also integrated safety features, providing developers with fine-grained controls to prevent high-risk actions. The model is now available in public preview. Find out more → https://goo.gle/3J029v6

Google DeepMind is renewing its partnership with European Bioinformatics Institute | EMBL-EBI on the AlphaFold Protein Structure Database, and we're marking it with an update 🎉 The database is now synchronised with the latest protein sequence data in UniProt. This is a crucial step to ensure AlphaFold coverage for newly discovered proteins and keeps the resource as current as possible. Plus, we’ve incorporated protein isoforms for the first time, providing a more complete view of proteome diversity and its importance for disease and drug design. This renewed partnership ensures that the global research community will continue to have open access to state-of-the-art protein structure predictions. Learning about work researchers around the world are conducting leveraging AlphaFold makes me hopeful about the potential future breakthroughs than can really help humanity. Cannot wait to see what the community builds with these new capabilities! In the meantime, read more about the partnership in the blog below ⬇️

Today, we’re sharing details about CodeMender: our new AI agent that automatically repairs serious code security flaws. 🛠️ Using the capabilities of Gemini Deep Think, it’s designed to help developers fix issues faster than ever before. But how can we trust AI to do this? Through rigorous verification. 🕵️ CodeMender uses advanced analysis to ensure the fixes are correct and prevents breakages in the wider code. It only sends fully vetted, high-quality solutions for human review. And it’s already created and submitted 72 high-quality fixes for serious security issues in major open-source projects. We hope to eventually release CodeMender as a tool for all developers, to enhance software safety for everyone. Find out more → https://lnkd.in/efppuFh2

How can AI enhance the creative process of a world-renowned industrial designer Ross Lovegrove? 🎨 Here’s how Lovegrove Studio and Modem used our latest models to develop a new kind of chair. 🪑 We fed a vast archive of Ross’ sketches to our text-to-image model to generate ideas rooted in his design language. And with Gemini, he was able to add more details, before bringing it to life using 3D printing. ↓