October Tech News: Reshaping Technology’s Future

October proved to be a watershed moment for technology. From next-generation language models and agentic assistants to AI-powered chips and creative tools, the past few weeks have seen a wave of groundbreaking announcements. Here are the stories that define the month.

OpenAI launches ChatGPT Atlas browser

OpenAI has launched a new AI-powered web browser called ChatGPT Atlas, initially available for macOS users. Support for Windows, iOS, and Android is expected soon. This browser integrates ChatGPT directly into the browsing experience, aiming to make web navigation more interactive and conversational. The launch reflects growing competition among tech companies to embed AI assistants into everyday tools, with similar moves from Google and Perplexity.

OpenAI releases Sora 2

OpenAI has released Sora 2, a groundbreaking video and audio generation model that significantly advances realism, physical accuracy, and controllability in AI-generated media. Built on the original Sora, this new version can simulate complex physical interactions and failures, follow multi-shot instructions, and generate synchronized soundscapes and dialogue.

It powers a new iOS social app called Sora, where users can create and remix content, including inserting themselves into scenes via “cameos.” The app emphasizes creativity over consumption, includes safety features for teens, and introduces natural language-controlled recommendation algorithms.

Windows 10 support ends

Windows 10 support officially ended on October 14, 2025, meaning Microsoft no longer provides technical assistance, feature updates, or security patches for the operating system. Users are encouraged to upgrade to Windows 11 for a more secure and efficient experience. If a device doesn’t meet Windows 11 requirements, users can either join the Extended Security Updates (ESU) program for temporary protection or purchase a new Windows 11-compatible PC.

Microsoft brings agentic AI to Windows 11

Microsoft is introducing Copilot Actions to Windows 11, a major step in its agentic AI strategy that transforms Copilot from a passive assistant into a proactive one capable of executing multi-step tasks like organizing files or replying to emails using natural language or voice commands.

Enabled via the “Hey Copilot” feature, this AI can automate workflows, interact with system settings, and operate independently within its own desktop environment. Microsoft also plans to replace the traditional search box with Copilot for faster access, integrating voice and text input.

Adobe launches AI Foundry

Adobe has launched Adobe AI Foundry, a new enterprise service that enables businesses to create custom generative AI models tailored to their brand and intellectual property. Built on Adobe’s Firefly models, which are trained exclusively on licensed data, these custom models can generate text, images, video, and even 3D scenes.

The Foundry service is priced based on usage and is designed to help companies personalize content across formats, languages, and seasons while maintaining brand consistency. Adobe emphasizes that these tools are meant to enhance, not replace, human creativity, positioning Firefly and Foundry as the next evolution in empowering storytellers and marketers.

Salesforce announces Agentforce 360

Salesforce has unveiled Agentforce 360, an upgraded AI agent platform designed to enhance enterprise productivity through smarter, more flexible automation. Announced ahead of its Dreamforce conference, the platform introduces tools like Agent Script for conditional AI behavior, Agentforce Builder for streamlined agent creation, and deeper integration with Slack, including a personalized Slackbot.

Powered by models from OpenAI, Anthropic, and Google Gemini, Agentforce 360 aims to improve reasoning and responsiveness in customer interactions.

Anthropic launches Claude Haiku 4.5

Anthropic has launched Claude Haiku 4.5, a smaller, faster, and more affordable AI model designed for users seeking quick and accurate responses. Despite its compact size, Haiku 4.5 rivals larger models like Claude Sonnet 4.5 and OpenAI’s GPT-5 in coding tasks, making it ideal for high-volume, real-time applications. The model complements Claude Sonnet 4.5 by handling subtasks in multi-step plans, offering businesses a scalable AI solution.

Google releases Veo 3.1

Google has released Veo 3.1, an upgraded version of its AI video generation model, now integrated into the Flow video editor, Gemini App, and Vertex/Gemini APIs. This update enhances realism and prompt adherence, introduces improved audio output, and adds granular editing tools like object insertion and future object removal.

Veo 3.1 includes some previous features, such as reference image-driven animation, frame-based clip generation, and video extension. Since Flow’s launch in May, users have created over 275 million videos, showcasing the model’s growing popularity and creative potential.

Apple M5 chip powers on-device AI

Apple has unveiled its new M5 chip, marking a major leap in AI and graphics performance across its product lineup, including the 14-inch MacBook Pro, iPad Pro, and Apple Vision Pro. Built on third-generation 3nm technology, the M5 features a 10-core GPU with Neural Accelerators in each core, delivering over 4x the peak GPU compute performance for AI compared to the M4. It also includes a faster 10-core CPU, a 16-core Neural Engine, and a 30% boost in unified memory bandwidth.

The M5 supports advanced features like ray tracing, spatial scene generation, and on-device large language model execution, while also contributing to Apple’s environmental goals through energy-efficient design.

Ant Group presents Ling-1T

Ant Group has announced the release and open-sourcing of Ling-1T, a trillion-parameter general-purpose large language model, expanding its Ling AI model family. The Ling family now includes three main series: Ling (non-thinking models), Ring (thinking models), and Ming (multimodal models).

Ling-1T delivers state-of-the-art performance in complex reasoning tasks, including code generation and competition-level mathematics. This follows the earlier release of Ring-1T-preview, the world’s first open-source trillion-parameter thinking model. Ant Group emphasizes its commitment to making Artificial General Intelligence (AGI) a public good through open and inclusive innovation.

October’s developments have set the stage for an exciting finish to 2025. Stay tuned as these innovations continue to unfold and shape the technological landscape.