Google Drive introduces AI-powered "Catch Me Up" summaries for recent file edits, comments, and updates, driven by Gemini AI. This feature provides concise overviews directly in the sidebar or per file, saving time for users managing collaborations. Accessible to select Workspace and AI plan subscribers, it highlights key changes like document edits and new comments. Summaries are available in English and will roll out over two weeks, though occasional inaccuracies may occur. The feature is enab

OpenAI enhances ChatGPT with cloud service integrations for business users, including Box, Dropbox, Google Drive, and more. It can now record and transcribe meetings, generate insights, and comply with organizational access controls. Additionally, a new Model Context Protocol expands research capabilities by leveraging third-party tools. These updates aim to make ChatGPT more office-friendly, competing with platforms like Zoom and Notion. Enterprise adoption is growing, reaching over 3 million s

Amazon plans to create a new team in its consumer division focused on autonomous AI, housed within Lab126. This hardware R&D group, known for developing Echo and Kindle, may aim to enhance robotics technology and improve warehouse robot functionalities. TechCrunch has contacted Amazon for further details.

Anysphere, the company behind AI coding assistant Cursor, raised $900 million at a $9.9 billion valuation in a round led by Thrive Capital. This is its third funding round in under a year, with ARR doubling every two months and now exceeding $500 million. Cursor offers tiered pricing for developers and recently introduced enterprise licenses, shifting from primarily individual subscriptions. Earlier this year, Anysphere rejected acquisition offers from OpenAI and others.

Google updated its Gemini 2.5 Pro preview model, improving performance in coding tasks and challenging benchmarks. Based on feedback, enhancements focus on style, structure, and creative response formats. The model is now available on AI Studio, Vertex AI, and the Gemini app.

OpenAI enhances ChatGPT for commercial users with cloud service integrations, meeting recording, and MCP support. New connectors for Dropbox, Box, SharePoint, OneDrive, and Google Drive allow retrieving information to answer queries. Features comply with organizational access controls and include transcription functionalities for meetings, creating timestamped notes and follow-ups. Deep research connectors (beta) for HubSpot, Linear, and Microsoft/Google tools enable comprehensive report generat

Reddit is suing Anthropic for using its data to train AI models without proper licensing, violating user agreements and exploiting content commercially. This marks the first major tech lawsuit over AI training data. Reddit has agreements with other AI firms like OpenAI and Google, ensuring terms protect user privacy. Reddit claims Anthropic ignored requests to stop scraping and seeks damages and an injunction.

Google is delaying the "Ask Photos" feature rollout by two weeks due to challenges in delays, quality, and user experience. This AI-powered tool, introduced at last year's Google I/O, lets users search their Google Photos library with natural language prompts. An improved version will be released soon.

Mistral AI launches Mistral Code Assistant to address compliance and security challenges in enterprise AI coding tools. The solution integrates Codestral, Devstral, and other models, offering fine-tuning on private code repositories and keeping code within enterprise boundaries. Features include RBAC, audit logs, and cloud or on-premises deployment options. Early adopters include Abanca, SNCF, and Capgemini. Despite competition from GitHub Copilot and OpenAI's Codex, Mistral targets the growing

Hugging Face unveiled SmolVLA, a compact yet powerful open AI robotics model that outperforms larger models in various environments. Trained on community-shared datasets with 450 million parameters, it runs on consumer-grade hardware like GPUs or MacBooks. SmolVLA supports asynchronous inference for faster responses and is part of Hugging Face's growing affordable robotics ecosystem. It joins other players like Nvidia and K-Scale Labs in the open robotics field.

Anthropic launched Claude Opus 4 and Sonnet 4, advanced LLMs with improved reasoning, tool use, and memory. Claude Opus 4 excels in coding benchmarks, achieving 72.5% on SWE-bench. These hybrid models support extended reasoning, simultaneous tool execution, and local file storage. They underwent rigorous safety testing under AI Safety Level 3 standards and reduce task shortcut tendencies by 65%. The release includes fully available Claude Code for JetBrains and VS Code.

Adobe launches Photoshop beta for Android, bringing core desktop features like layers, AI tools, and generative fill powered by Firefly. Available in select regions, the app supports PSD files, advanced selection tools, and requires Android 11+ with 6GB RAM. Tutorials help users master its functionalities, following the iPhone version's release in February 2025. Amid AI advancements, Adobe addresses concerns about using customer content for training models without consent.

Google's NotebookLM now supports public sharing via simple links, enabling read-only access to AI-generated content like summaries and FAQs. This update enhances collaboration for study guides, project briefs, and more, allowing viewers to interact without editing original materials. The feature aims to integrate generative AI into productivity workflows, making complex information easier to share and understand.

OpenAI enhances Codex by adding internet access, expanding availability to ChatGPT Plus users, and introducing a voice command feature. The update allows Codex to install dependencies, run tests, and modify existing pull requests, improving efficiency and accessibility while maintaining security safeguards. These improvements position Codex as a more integral tool in development workflows.

Google launched AI Edge Gallery, an experimental app that runs AI models offline on smartphones without cloud or internet. Available under Apache 2.0 license, it initially supports Android with iOS coming soon. Key features include AI Chat, Image Query, and Prompt Lab. Users can download models like Gemma 3n, with sizes ranging from 554 MB to 4.4 GB. Performance is impressive, especially for smaller models, offering privacy benefits by processing data locally. Multimodal capabilities work well,

Google launched AI Edge Gallery, an Android app enabling users to run AI models on their devices without cloud connectivity. Currently featuring 18 models from Google, Microsoft, Alibaba, and others, it offers key features like AI Chat, Ask Image, and Prompt Lab. This initiative advances mobile AI despite device processing limitations.

IBM acquired Seek AI, a platform enabling natural language querying of enterprise data. Seek's tech will integrate into IBM's Watsonx AI Labs, an NYC-based AI accelerator. Founded in 2021, Seek raised $10 million and offers tools for data utilization via chatbot-like interfaces. The acquisition supports IBM's growth in enterprise AI, following strong Q1 earnings driven by AI demand. Watsonx AI Labs aims to collaborate with universities and support startups, leveraging IBM's resources and talent

Claude 4, Anthropic's latest AI model, is praised as the most advanced coding tool. This article explores five methods to integrate Claude 4 into workflows: using the Claude AI web app for GitHub repo integration, setting up Claude Code for direct project enhancement, automating GitHub workflows with @claude commands, integrating with VSCode for an intelligent IDE assistant, and leveraging the Python SDK for API-based access. These tools boost productivity and streamline development tasks.

Character.AI launched new multimedia features including AvatarFX, Scenes, and Streams, enabling users to create and share AI-generated videos. While expanding its offerings beyond text chats, the platform aims to balance creativity with safety, adding watermarks and preventing real-person uploads to curb misuse. However, concerns remain about potential abuse and ethical challenges despite these safeguards.

Microsoft Bing introduces Bing Video Creator, a tool using OpenAI's Sora model to generate videos from text prompts. Initially for paying customers, this is now free in the Bing app. Currently mobile-only, video creation may take hours despite targeting minutes. Users get ten free clips after signing in, then pay via Microsoft Reward points earned through searches or shopping. Videos are initially 5 seconds, vertical 9:16 ratio, with horizontal support coming soon.