Anthropic's First Ever Chrome Extension🌐

From safer AI assistants in your browser to lightning-fast models and intuitive video creation apps, the AI world is moving quickly:

• Anthropic’s Claude Chrome Extension brings AI directly into your browser to manage emails, calendars, and more—with robust safety features and new defenses against prompt injection attacks.

• Google Vids makes professional video creation effortless inside Google Workspace, with AI storyboards, templates, avatars, and real-time collaboration.

• NVIDIA’s Nemotron Nano 9B V2 debuts as a high-speed hybrid Transformer model, delivering six times faster inference and massive context handling—ideal for AI agents and enterprise workloads.

Stay tuned as these tools redefine productivity, creativity, and the speed of AI innovation.

Anthropic Unveils Claude Chrome Extension

Anthropic is piloting a Chrome extension that lets trusted users instruct Claude to take actions directly within their browser, offering practical benefits like managing calendars, handling emails, and automating routine tasks, while prioritizing safety through measures such as action confirmations, site-level permissions, restrictions on high-risk websites, and advanced attack detection; thanks to robust mitigations—including new prompt injection defenses—the attack success rate was reduced from 23.6% to 11.2% in testing, with some browser-specific threats cut to zero, but wider release depends on feedback from real-world users to further strengthen security before broader availability.

Make Videos Fast with Google Vids

Google Vids is an AI-powered video creation app integrated within Google Workspace that enables users to create engaging and professional videos easily, even without prior editing experience. It features AI-generated storyboards, customizable templates, voiceover options with Google's AI voice technology, and access to extensive stock media and fonts. Users can start video creation from simple prompts or files, collaborate in real time, and personalize videos with AI avatars, animated images, and automatic editing tools. The platform supports generating short video clips from text prompts using advanced AI models like Veo 3, making video production quick and dynamic for training, marketing, product demos, and announcements.

Nemotron Nano 9B V2: Fast Hybrid Transformer AI

NVIDIA-Nemotron-Nano-9B-v2 is a state-of-the-art large language model developed by NVIDIA, designed to excel in both reasoning and non-reasoning tasks. It features a hybrid Mamba-Transformer architecture, primarily composed of Mamba-2 and MLP layers, with only four attention layers, enabling faster inference and high accuracy. The model supports runtime "thinking" budget control, allowing users to specify how much internal reasoning it performs during inference, which helps balance accuracy and response time. Pretrained on a massive dataset of approximately 20 trillion tokens, it supports multiple languages including English, German, Spanish, French, Italian, and Japanese, as well as programming languages. The model can handle very long context windows of up to 128,000 tokens on a single NVIDIA A10G GPU, achieving up to six times faster throughput compared to similar-sized models while maintaining or surpassing their accuracy. It is well-suited for AI agents, chatbots, retrieval-augmented generation systems, and typical instruction-following tasks, and is commercially available under the NVIDIA Open Model License.

Simplify your content creation with powerful AI tools! Trim silences, remove noise, merge and loop audio, and even remove video backgrounds — all with just a few clicks. Join over 15,000 creators who save time and produce professional-quality videos, podcasts, and more effortlessly.

Try Now

Learn Claude Code

Top AI Products from this week

• Webvizio for AI Coding Agents – Convert non-technical feedback and bug reports into ready-made code in seconds. Webvizio captures product feedback, automatically gathers all technical data, and helps your team quickly resolve dev tasks using AI coding tools like Cursor or Windsurf.

• Rube – Rube is a universal MCP that enables you to take actions across 600+ applications from within your AI chat, be it Cursor, Claude, VS Code, etc. It manages authentication and tool selection seamlessly and securely within the chat.

• Roark – Build voice agents you can trust. Roark tracks call metrics, runs evaluations, and stress-tests your agent with simulated callers across accents, languages, and speaking styles. Failed calls become tests - giving you visibility and continuous improvement.

• Conductor – Conductor lets you run a bunch of Claude Codes all at once, on your Mac. Each Claude gets an isolated copy of your codebase. See at a glance what they're working on, then review and merge their changes.

• Hex Notebook Agent – Let Hex's agent abstract away the most tedious parts of data work, so you focus on the new, the novel, and the gnarly. Hex brings the best of Claude Sonnet 4 directly to our powerful notebooks where thousands of data teams are already working.

• TaskGPT Mobile - helps development teams find security vulnerabilities before deployment.

This week in AI

• UE8M0 FP8 Scaling AI Efficiently - DeepSeek’s UE8M0 FP8 boosts AI scaling with lower memory use, high throughput & domestic chip alignment—balancing accuracy, efficiency, and hardware portability.

• Character.AI boom - 28M users, 2B+ monthly chat mins, 20%+ more engagement, fine-tuned open-source models deliver fun, coherent AI stories, driving huge user joy and growth.

• Gemini 2.5 Flash Image - Gemini 2.5 Flash Image blends, edits, and fuses multiple images with natural language. It keeps character consistency, uses world knowledge, enables precise edits, and supports creative workflows.

• Asta AI for Science - Ai2’s Asta is an open-source AI ecosystem aiding science with agents that find papers, summarize literature, analyze data, and benchmark AI tools for trust and rigor.

• HeyGen Avatar IV - HeyGen Digital Twin, powered by Avatar IV, creates lifelike full-body avatars mirroring gestures, expressions, and voice, enabling stress-free, realistic video creation instantly.

Paper of The Day

An Agentic AI Framework for Enhancing Scam Intelligence in Digital Payments" presents a novel AI system deployed in Google Pay India that uses conversational agents powered by Google’s Gemini models to interview users targeted by scams. This agent collects detailed user scam reports through empathetic, dynamic dialogue, generating rich, structured scam intelligence. An information extractor converts these transcripts into actionable data, significantly improving scam detection by 21% and speeding enforcement. The framework emphasizes safety, user privacy, and scalability, offering a generalizable approach for combating sophisticated scams and other abuse in digital payments and beyond.

To read the whole paper, go to here.