Meta has opened Muse Voice, the voice call feature of its AI agent Muse, to every user in the United States. It works on mobile, web, and Mac, and lets people talk face to face with animated avatars. However, Muse cannot actually see the user's face.
Amazon Web Services (AWS) has officially launched Web Answer, a feature that AI agents can use. After sending one question, users can get a complete answer with sources, based on the latest web information, all at once.
Meta has opened Muse Voice, a feature that lets people talk with its AI agent Muse by voice or video, to all users in the United States. The feature was first introduced at Meta's developer event Connect in September this year.
ElevenLabs has released the ElevenAgents SDK, a tool for Unity game development. It connects game characters to voice agents (AI programs that handle conversation), so when a player walks up to a character and speaks into a microphone, the character replies in the agent's voice.
AI agents handle tasks like searching, querying databases, and writing and running code, and each of these steps needs a CPU (the central processing unit of a computer). NVIDIA says its new Vera CPU delivers both high speed and high throughput per unit of power in a single chip.
AI agents, which are automated programs that do work on behalf of users, now move around the web for millions of people. It has become hard for websites to tell who the real user is. This article argues that websites need new ways to recognize the user behind an agent and hold them accountable.
The optimized engine of Amazon OpenSearch Service can now receive AI agent trace data. Users can view trace data, which records workflow steps one by one, and logs in one place. They can follow a request or agent task from start to finish to find steps that failed or slowed down.
The DeepLearning.AI channel released a video on how to balance latency (the time it takes to respond) and reliability in voice AI. It covers how to build voice agents (AI programs that carry out tasks on their own) that keep up with a conversation while handling tasks dependably.
AI coding tools can now build features of tens of thousands of lines at once, making it hard for people to review code by reading it line by line. Microsoft Research introduced PDD Kit, a tool that uses mathematical proofs to check whether code behaves according to its specification.
Kakao's Applied AI Model team has released Kanana-2-v-decide as open source. It is a multimodal decision-making model for AI agents, which are AI systems that judge and handle tasks on their own. Multimodal means handling several kinds of information together, such as text and photos. The introduction says it covers tasks such as sorting customer inquiries and processing product photos.
Kimi has officially launched a hosted agent service that lets businesses and developers run long-running AI tasks and multiple AI agents working together in the cloud. The service became a full product after its beta in September 2026.
Mistral has expanded Mistral Studio, its enterprise AI platform. Companies can use AI agents (AI programs that carry out tasks on their own) connected to their company data and systems, along with workflows and apps, deployed in the cloud or on their own servers. Mistral also introduced its new large model, Mistral Large 4.
This video explains how to create your first AI agent, which divides up work on its own, from the perspective of an e-commerce CEO. It introduces an AI agent as a single-page job description (a document listing the duties to be handled) rather than a complicated program. The topic is how to assign your first three AI agents to handle sales, email, and marketing.
Postman's desktop app automatically upgraded and changed the settings of Codex and Claude Code, even though users did not click any confirmation prompt. The app's installer also wrote the hook trust records, not the user. According to the analysis, the boundary that protects security is in the settings files, not in the confirmation window.
Meta has acquired Conductor, a startup that lets developers run and manage several AI coding tools at the same time from one screen. As AI increasingly writes code for developers, Meta appears to be strengthening its position in the developer tools market. The deal terms, including the purchase price, have not been disclosed.
OpenAI has introduced GPT-6 Cyber, a cybersecurity model. It is designed to find and fix vulnerabilities that could be used in attacks, and it was trained to act only within permitted boundaries.
This AWS weekly roundup covers the week's news. The main items are that OpenAI's GPT-6.1 Terra model is now generally available on Amazon Bedrock, and that AWS Step Functions workflows can now be called as tools by AI agents.