A collection of the most interesting AI news. # June 2025 ## Google - [**Gemma 3n Released**](https://ai.google.dev/gemma/docs/releases): Google released Gemma 3n, a lightweight open-source language model optimized for everyday devices and now featuring audio input. - [**Search Live Mode**](https://searchengineland.com/google-launches-search-live-with-talk-and-listen-within-google-app-457224): Google rolled out "Search Live" (in the US only), a talk and listen conversational mode within the Google Search app, allowing back-and-forth voice interactions and real-time link exploration. - [**A2A Joins Linux Foundation**](https://www.linuxfoundation.org/press/linux-foundation-launches-the-agent2agent-protocol-project-to-enable-secure-intelligent-communication-between-ai-agents): Google's Agent2Agent (A2A) protocol, designed for secure agent-to-agent communication and collaboration, officially joined the Linux Foundation, aiming for open-source neutrality and broader adoption. - [**Gemini CLI Released**](https://blog.google/technology/developers/introducing-gemini-cli-open-source-ai-agent/): Google unveiled Gemini CLI, a new open-source agentic AI tool runnable locally from terminals, allowing developers to interact with Gemini models using natural language for coding and other tasks. - **[New Agent Mode in Gemini Code Assist](https://cloud.google.com/gemini/docs/release-notes) (VS Code)**: Gemini Code Assist introduced a new agent mode in preview for VS Code, enabling interactive reviews, multi-file editing, and full project context for complex, multi-step coding tasks. - **[Doppl Released in the US](https://labs.google/doppl/)**: Google launched Doppl in the U.S., an app designed for virtual try-on, allowing users to see how outfits would look on them without physically trying them on. - **[Google AI Ultra for Business:](https://workspace.google.com/blog/product-announcements/google-ai-ultra-for-business)** Google announced Google AI Ultra for Business, a new add-on for Workspace aimed at boosting productivity and creativity for specialized teams. - YouTube has announced the integration of Google DeepMind's latest and most advanced AI video generation model, [Veo 3, into YouTube Shorts](https://www.theverge.com/news/689474/youtube-veo-3-ai-videos-shorts), expected later this summer. ## Microsoft - **[Mu](https://blogs.windows.com/windowsexperience/2025/06/23/introducing-mu-language-model-and-how-it-enabled-the-agent-in-windows-settings/) - An LLM Enabling Agents on Windows Computers:** Microsoft introduced the Mu language model, a new on-device small language model. It's optimized for NPUs and edge devices. - [**GitHub Copilot Coding Agent for Business Users:**](https://github.blog/changelog/2025-06-24-github-copilot-coding-agent-is-now-available-for-copilot-business-users/) The GitHub Copilot coding agent became available for Copilot Business subscribers, with continued improvements to its agent capabilities for tasks like generating pull requests and revising code. ## OpenAI - [Open-Source Customer Support Agent Demo](https://venturebeat.com/programming-development/openai-open-sourced-a-new-customer-service-agent-framework-learn-more-about-its-growing-enterprise-strategy/): OpenAI released an open-source airline customer service agent as a demonstration of building agents and multi-agent workflows using the Agents SDK. - [GPT-5 Announced](https://explodingtopics.com/blog/new-chatgpt-release-date) (expected Summer 2025): Sam Altman, CEO of OpenAI, hinted at the upcoming release of GPT-5 in Summer 2025, signaling a new leap in generative AI. ## Others - **[MIT Study Explores Impact of ChatGPT on Cognitive Engagement](https://www.media.mit.edu/publications/your-brain-on-chatgpt/)**: Over-reliance on generative might suppress cognitive engagement and memory, potentially leading to reduced neural activity. However, for experts, AI can serve as a powerful augmentative tool, freeing them from mundane tasks to focus on higher-level critical thinking, creativity, and strategic decision-making. - [**Anthropic's Claude Supports MCP Server](https://www.anthropic.com/news/claude-code-remote-mcp):** Claude Code announced support for remote Model Context Protocol (MCP) servers, allowing integration with various tools and data sources for personalized coding experiences without local server management. - [**IBM Launches Industry-First Platform for Agentic AI Governance and Security](https://newsroom.ibm.com/2025-06-18-ibm-introduces-industry-first-software-to-unify-agentic-governance-and-security):** IBM introduced a new software stack unifying watsonx.governance and Guardium AI Security platforms to provide centralized oversight for autonomous AI systems in enterprises. - **[Postman's AI-Ready APIs Initiative](https://www.postman.com/ai/ai-ready-apis/):** Postman's AI-Ready APIs initiative focuses on helping organizations build strong API foundations—emphasizing well-structured, consistent, and reliable APIs—to support the growing demands of AI agents. - [**OpenAI o3-Pro**](https://lnkd.in/d6Rp4aQH): OpenAI released o3-Pro, its most advanced reasoning model to date. It can analyze contracts, market data, and perform complex decision-making tasks traditionally handled by consultants. - [**Google Gemini CLI**](https://lnkd.in/dufzShqZ): Google launched a command-line interface for Gemini, enabling natural language commands in terminal workflows—e.g., "deploy my app to production" triggers full deployment pipelines. - [**Mistral Vibe Coding Assistant**](https://lnkd.in/d2hf6YqV): Mistral introduced Vibe, a highly customizable enterprise-grade coding assistant that surpasses GitHub Copilot in debugging and code generation for production use. - [**Midjourney V1 Video**](https://lnkd.in/d9kuueyZ): Midjourney launched its first video model, generating cinematic-quality clips from text prompts. Users can reframe scenes while retaining stylistic fidelity and visual quality. - [**Higgsfield Soul Model**](https://lnkd.in/dss-j5Ag): Higgsfield unveiled Soul, an ultra-realistic AI photo generator with 50+ curated presets, targeting creators and brands aiming for fashion-grade content realism. - [**HeyGen Avatar IV**](https://lnkd.in/dZFxj3DP): HeyGen released Avatar IV, a lifelike digital avatar generator with 99.7% voice and visual cloning accuracy, micro-expressions, and full-body motion control. - [**ElevenLabs 11ai Voice Assistant**](https://lnkd.in/dg7ThgXx): ElevenLabs launched a real-time voice assistant capable of natural human-like conversation with 150ms latency, emotional tone, and live interruption handling. - [**Runner H Agent**](https://lnkd.in/d3Q5XW6j): Runner introduced H Agent, an autonomous AI developer that can build full-stack applications from a single prompt—recently demonstrated by building an e-commerce site in 47 minutes. - [**MiniMax M1 Model & AI Agent**](https://lnkd.in/dYAtibZh): China’s MiniMax released the M1 model, claiming GPT-4 parity at 10x lower cost, along with agents capable of autonomously building apps, running code, and generating presentations. - [**Claude Code MCP Servers**](https://lnkd.in/dWvMfqNN): Anthropic’s Claude Code now supports Model Context Protocol (MCP) servers, enabling direct access to dev environments for codebase reading, test execution, and GitHub integration. # May 2025 - [Google I/O Developer Conference 2025](https://io.google/2025/) - Google [Gemini 2.5 Pro](https://deepmind.google/models/gemini/pro/) gets Deep Think enhanced reasoning, native audio output and [live API](https://ai.google.dev/gemini-api/docs/live). It also excels in advanced coding. It is a natively multimodal with 1-mio taken context window. - [Google 2.5 Flash](https://deepmind.google/models/gemini/flash/) model is released and designed for speed and low-cost. - [Google AI Mode](https://blog.google/products/search/ai-mode-search/) brings AI to search. - [AI Shopping](https://blog.google/products/shopping/google-shopping-ai-mode-virtual-try-on-update/) will also be possible in the AI Mode. - Google Chrome gets [Chrome Gemini integration](https://gemini.google/overview/gemini-in-chrome/?hl=en). - Google brings [Gemini AI to IoT devices](https://blog.google/products/android/gemini-watches-cars-tvs-xr/) like smartwatches, cars, TVs, and XR headsets. - Google releases an AI-powered video tool called [Flow](https://labs.google/flow/about). - [Google Stich](https://stitch.withgoogle.com/) helps you to design at the speed of AI. - Google [Veo 3](https://deepmind.google/models/veo/) in Flow brought voice to AI video. - [Imagen4](https://deepmind.google/models/imagen/) is new Google's image generator. - The goal of [Project Astra](https://deepmind.google/models/project-astra/) is to build the universal AI assistant. - [Google Beam](https://blog.google/technology/research/project-starline-google-beam-update/) demonstrated an AI-first 3D video communication platform and brought real-time translations to Google Meet. - [Android XR](https://www.android.com/xr/) is an AI-powered OS coming to headsets and glasses. - [Project Mariner ](https://deepmind.google/models/project-mariner/), a web browsing agent, can observe your web browser and then plan and act on complex tasks. - [Vertex AI](https://cloud.google.com/vertex-ai), a ML platform that lets you train and deploy ML models and AI applications, comes with a prompt library. - [Jules](https://jules.google/) is Google's new asynchronous coding agent. - [Google AI Studio](https://aistudio.google.com/) gets native code editor and GenAI SDK. - [NotebookLM](https://notebooklm.google/), your research and thinking partner, gets a mobile app. - [Gemma 3n](https://deepmind.google/models/gemma/gemma-3n/) can now run on mobile phones. - [AlphaEvolve](https://deepmind.google/discover/blog/alphaevolve-a-gemini-powered-coding-agent-for-designing-advanced-algorithms/) is a Gemini-powered coding agent for designing advanced algorithms. - [Fire Sat](https://sites.research.google/gr/wildfires/firesat/): As part of its disaster response efforts, Google unveiled Fire Sat, an AI-powered system for early wildfire detection using satellite imagery. - Google announces [AI Ultra Plan - Pro subscription](https://blog.google/products/google-one/google-ai-ultra/) for $250 / month with higher limits and access to experimental features. - [Microsoft Build 2025 Conference](https://news.microsoft.com/build-2025/) - [Windows AI Foundry](https://developer.microsoft.com/en-us/windows/ai/) is now a unified platform that supports complete AI developer lifecycle from model selection, optimization, tuning, and deployment. - New models are added to Azure, including xAI’s Grok3 and Grok3 mini. - Azure AI Agent Service is now available for building enterprise-grade AI agents. - The Model Context Protocol (MCP) aims to be the "HTTP of AI agents" for cross-platform communication. - [Microsoft Discovery](https://azure.microsoft.com/en-us/blog/transforming-rd-with-agentic-ai-introducing-microsoft-discovery/) is an enterprise AI platform that helps accelerate research and discovery. - [Microsoft Copilot Studio](https://www.microsoft.com/en-us/microsoft-copilot/microsoft-copilot-studio) gets multi-agent orchestration, maker, and more. - [Agentic DevOps](https://azure.microsoft.com/en-us/blog/agentic-devops-evolving-software-development-with-github-copilot-and-microsoft-azure/) - GitHub introduced DevOps with AI Agents. - GitHub is launching an AI coding agent directly embedded into[ GitHub Copilot](https://github.blog/changelog/2025-05-19-github-copilot-coding-agent-in-public-preview/). - [Power Apps](https://www.microsoft.com/en-us/power-platform/blog/power-apps/reimagining-human-agent-collaboration-for-a-new-era-of-app-development-with-microsoft-power-apps/) gets updates to accelerate app creation and govern agentic experiences. - [Microsoft Fabric](https://blog.fabric.microsoft.com/en-us/blog/digital-twin-builder-in-microsoft-fabric-real-time-intelligence-revolutionizing-digital-twin-creation-and-management) is a data platform for a faster AI transformation. It also democratizes and scales digital twin creation and management. - [NLWeb](https://news.microsoft.com/source/features/company-news/introducing-nlweb-bringing-conversational-interfaces-directly-to-the-web/) helps you turn websites into AI apps. With the new it's much easier to add AI chatbots to websites. - Windows File explorer will get [AI Actions](https://www.theverge.com/news/670251/microsoft-windows-11-ai-actions-file-explorer-context-menu). - PDFs can be [translated directly in MS Edge](https://www.microsoft.com/en-us/edge/features/pdf-translation?ch=1&form=MA13FJ). - **OpenAI** - OpenAI is launching a new AI coding assistant called [Codex](https://openai.com/index/introducing-codex/). - [Operator AI Assistant](https://openai.com/index/introducing-operator/) is now capable of handling various online tasks, such as ordering groceries and processing ticket purchases. - OpenAI’s flagship [GPT-4.1](https://openai.com/index/gpt-4-1/) and 4.1.-mini models are now available in ChatGPT. - You can now [connect GitHib to ChatGPT deep research](https://help.openai.com/en/articles/11145903-connecting-github-to-chatgpt-deep-research). - OpenAI buys iPhone designer Jony Ive device startup [io](https://openai.com/sam-and-jony/) for $6.5 billion. They aim to develop screen-free, pocket-sized AI companion devices designed to deeply understand and anticipate users' needs, potentially replacing traditional screen-based interactions. - OpenAI to [acquire](https://www.bloomberg.com/news/articles/2025-05-06/openai-reaches-agreement-to-buy-startup-windsurf-for-3-billion) startup [Windsurf](https://windsurf.com/editor) for $3 billion. - ChatGPT in May 2025 surpassed Wikipedia in monthly users. - OpenAI launched PDF exports for deep research reports. - **Anthropic** releases [Claude Opus 4](https://www.anthropic.com/claude/opus), a hybrid reasoning model that pushes the frontier for coding and AI agents, featuring a 200K context window. It can code for hours autonomously. - **Mistral** launches [Devstral](https://mistral.ai/news/devstral) Small 24B, a new open-source LLM for coding, and [Medium 3](https://mistral.ai/news/mistral-medium-3), that delivers state-of-the-art performance at 8x lower costs. - You can now add [Perplexity AI to WhatsApp](https://www.perplexity.ai/changelog/what-we-shipped-may-2nd). You can do the same with [ChatGPT](https://help.openai.com/en/articles/10193193-1-800-chatgpt-calling-and-messaging-chatgpt-with-your-phone). - **Hugging face** [Open Computer Agent (OCA)](https://huggingface.co/spaces/smolagents/computer-agent) that you can prompt to complete tasks, the same way as [OpenAI's Operator](https://openai.com/index/introducing-operator/). - **Notion** released [AI Meeting Notes](https://www.notion.com/product/ai-meeting-notes), where you can capture every idea, decision, and next step in Notion, and your team can put them to work right away. - **Audible** empowering publishers with AI audiobook tools for [narration and translation](https://www.audible.com/about/newsroom/audible-expands-catalog-with-ai-narration-and-translation-for-publishers). - **Alibaba** releases [Qwen3](https://chat.qwen.ai/) and Web Coder. - [Saudi Arabia HUMAIN](https://www.pif.gov.sa/en/news-and-insights/press-releases/2025/hrh-crown-prince-launches-humain-as-global-ai-powerhouse/): Saudi Arabia launched HUMAIN, a state-backed AI company under the Public Investment Fund, aiming to position them as a global leader in AI innovation and infrastructure.