IBM and Microsoft have launched Imagine Enterprise, a platform for building and governing AI agents that assist with industry-specific work. IBM Consulting contributes more than 6,000 prebuilt agents and other assets, while Microsoft supplies cloud, data and security technology. Engineers from both companies will help qualifying customers move deployments into production.

Source

Gemini 4 Carbon, Google’s cloud-based AI model for complex coding, security and research tasks, is now available to paying developers and enterprises after an initial release to cybersecurity defenders. Its introductory API price matches Gemini 4 Argon at $2 per million input tokens and $10 per million output tokens. Google says Carbon is designed for harder, longer-running tasks, while Argon remains its recommendation for most software engineering and knowledge work.

Source

OpenAI has launched Businesses in ChatGPT in beta, giving companies a way to manage how they appear in the online AI service. The new Business Hub lets eligible businesses claim profiles, update product information and track engagement. Eligible US ecommerce and online service businesses can also launch Business Agents to answer customer questions.

Source

Grok Build, an online AI service for creating apps from natural-language instructions, left thousands of apps accessible without signing in, according to security firm Claroc. The exposed material included confidential sales documents from SpaceXSI, the company behind Grok.

Source

Anthropic’s Claude is an AI assistant available through the web, desktop and mobile apps. Its new Claude Meetings feature joins video calls to draft plans and work on tasks as people talk. The company also launched collaborative whiteboards and spreadsheets, and made Claude Data generally available.

Source

GitHub Copilot, an AI coding assistant available in development tools and online, now offers Google’s Gemini 4 Carbon model. GitHub says the model is designed for complex coding, reasoning, agent tasks, and security-focused work. Its availability is rolling out gradually, and usage is billed at provider list pricing.

Source

Microsoft Research has shown that editing fewer than 10 dormant neurons can turn an AI agent model into a sleeper agent that runs an attacker’s command when it encounters a trigger, such as a date change. The technique, called Lazarus, requires no model retraining and needs only inference-level compute. The modified model passed standard evaluations, raising concerns for agents that operate unattended with system access. Microsoft says it has not observed threat actors using the technique, and that AI agents can help detect the anomalous weight edits.

Source

Microsoft Foundry, a cloud platform for building AI applications, now offers Anthropic’s Claude Fable 5.2 and Mythos 5.2 models for technical work. The recently launched models are hosted by Anthropic through Foundry, with Mythos access restricted to verified organizations.

Source

Microsoft and IBM have launched Imagine Enterprise, a platform for building and managing AI agents that carry out tasks across business workflows. It combines Microsoft’s cloud and security services with IBM consulting tools and a library of more than 6,000 prebuilt agents and other assets. The offering targets regulated industries seeking to move AI projects into production.

Source

AMD has launched EPYC 6, its Venice server processor family, with models designed for AI agent workloads and other data center applications. The 10-chip lineup ranges from high-frequency processors to 256-core parts intended to run many agents concurrently. AMD says servers with the new chips are shipping, with Microsoft Azure and Oracle Cloud Infrastructure instances coming soon.

Source

Claude, Anthropic’s AI assistant, is moving further into online office productivity with three new artifact types: Sheets, Whiteboards and Data. The additions bring Anthropic closer to competing with Microsoft 365 and Google Workspace. The company also plans to add Claude Meetings, while expanding integrations that let users work with other software inside Claude.

Source

Apple’s Mac Studio is a desktop workstation that can run AI models locally using its M5 Ultra chip and unified memory. New benchmarks show that a 256 GB configuration completed an agentic AI workload in less than half the time of a DGX Spark-like system, but larger models exposed a crucial software gap. For those models, the choice of inference engine—the software that runs them—can determine whether the Mac or a two-system Nvidia cluster wins.

Source

ChatGPT regained ground among six major chatbots, increasing its share of sessions from 80.42% in June to 81.2% in September 2026, according to a Graphite analysis of Similarweb data.

Source

Matthew Berman reviews Anthropic’s Claude Fable 5.2 on coding, game development and graphics projects. His demonstrations include a full Microsoft Excel clone, a Windows video editor and playable games built from short prompts. He reports strong autonomous coding performance, but some projects consumed billions of tokens and would have cost thousands of dollars at API rates.

Source

Claude is Anthropic’s AI service for coding, research, and other technical work. Its new Claude Fable 5.2 and Claude Mythos 5.2 models generate output up to 30% faster than their 5.1 predecessors, according to Anthropic, and cost $8 per million input tokens and $40 per million output tokens. Fable is generally available, while direct Mythos access is restricted to approved cybersecurity and life sciences users.

Source

Anthropic’s Claude models help developers tackle coding and reasoning tasks through cloud services. Claude Fable 5.2 is now available on Amazon Bedrock and Claude Platform on AWS, alongside the restricted Claude Mythos 5.2. Organizations adopting Fable should check its data-retention settings and test existing integrations before migrating.

Source

Red Hat OpenShift manages containerized applications and virtual machines across data centers and clouds. The newly available OpenShift 5 adds AI-assisted troubleshooting and upgrade checks, while forthcoming Red Hat Enterprise Linux (RHEL) updates add tamper protection and AI administration tools. Autonomous troubleshooting remains a technology preview and requires human approval for changes.

Source