Latest Briefings
HTTPS Ecosystem Retires 11 Legacy Domain Validation Methods by 2028
The CA/Browser Forum and Chrome Root Program are phasing out email, phone, and reverse-lookup domain validation methods in favor of automated, cryptographically…
Google Project Zero Chains Pixel 9 Sandbox Escape via BigWave Kernel Driver
A Project Zero researcher found three bugs in the Pixel 9's BigWave AV1 hardware driver, one of which enables a full mediacodec…
Google Project Zero Found Nine Bypasses in Windows Administrator Protection
A Google Project Zero researcher identified nine separate vulnerabilities in Windows 11's new Administrator Protection feature before its release, all of which…
Project Zero Digs Into GetProcessHandleFromHwnd: Bugs, History, and Inaccurate Docs
A Google Project Zero researcher traces the evolution of a little-known Windows API across Vista, Windows 7, and Windows 10, exposing documentation…
Google DeepMind Proposes Cognitive Framework for Measuring AGI Progress
A new taxonomy from Google DeepMind identifies 10 cognitive abilities as benchmarks for general AI, backed by a $200,000 Kaggle hackathon to…
Google DeepMind Launches Gemini 3.5 Flash, Targeting Agentic AI Workloads
Google DeepMind has released Gemini 3.5 Flash, a model designed for long-horizon agentic tasks and complex coding workflows, claiming benchmark leads over…
Google DeepMind Releases Toolkit to Measure AI Manipulation Capabilities
DeepMind has published the first empirically validated framework for measuring harmful AI manipulation, drawing on nine studies with over 10,000 participants across…
Google DeepMind Launches Gemini 3.1 Flash Live with SynthID Watermarking
Google's latest real-time audio model brings lower latency, stronger reasoning benchmarks, and mandatory SynthID watermarking on all generated audio to curb AI-driven…
Google DeepMind Reimagines the Mouse Pointer with Gemini AI Context Awareness
Google DeepMind has outlined four design principles and experimental demos for an AI-powered pointer that uses Gemini to understand visual and semantic…
Google DeepMind Launches Gemini 3.1 Flash TTS With Audio Tags and SynthID Watermarking
Google DeepMind's latest text-to-speech model brings granular vocal control via natural language audio tags, native multi-speaker dialogue, and mandatory SynthID watermarking across…
Google DeepMind’s Decoupled DiLoCo Trains LLMs Across Data Centers with Minimal Bandwidth
Google DeepMind has published research on Decoupled DiLoCo, a distributed training architecture that isolates hardware failures to separate compute islands and dramatically…
AlphaEvolve Cuts DNA Sequencing Errors by 30% in DeepConsensus Upgrade
Google DeepMind's AlphaEvolve coding agent improved the DeepConsensus DNA error-correction model, reducing variant detection errors by 30% and lowering sequencing costs for…