Anthropic has unveiled Claude Duet, a new artificial intelligence system designed to operate as an active personal agent 24 hours a day. This feature aims to assist users in organizing their routines and optimizing available time.
The announcement was made this Wednesday, October 21st. The functionality of Claude Duet shares similarities with existing features, such as the ChatGPT Dots function, accessible in Brazil, Google's former Gemini Spark (now located in the Tasks section), and Meta's Muse, although the latter does not have a planned launch in the country.
These competing tools aim to generate summaries and offer suggestions based on users' digital activities. Initially, the Claude Duet technology is only available in beta version, targeted at the research sector, with a primary focus on professionals and small businesses. It utilizes the Claude Opus 5.5 base and can be accessed through a standalone application for Android and iPhone (iOS), as well as the claude.ai website.
With the launch of Claude Duet, Anthropic enters the segment of continuously operating agentic AIs, focusing on executing autonomous actions in daily life. The proposal is to integrate the model into the productivity applications and software most used by the user, enabling the collection of relevant daily emails, locating times in the calendar, performing financial projections, and issuing recommendations based on performed activities.
Although generative AIs can perform similar tasks, especially in their most advanced versions, this generally requires the input of very specific and detailed commands (prompts), subject to restrictions depending on the analyzed platforms. In the case of Duet, this process becomes automatic, resulting in simplified routine interactions, effectively functioning as a personal assistant.
The intention is for Duet to help manage individual activities. Furthermore, Anthropic mentioned the possibility of a virtual machine, following the Meta model, and promised independent operation via its own browser. Despite being more autonomous than traditional chatbots, Claude Duet maintains the responsibility for complex decisions with the user, who must approve payments, confirm message sending, and validate configuration changes, for example.
Considering that an assistant intended for professionals and small businesses may handle highly confidential data, Anthropic has implemented mechanisms to identify and analyze sensitive information. The agent is programmed to refuse sharing credit card numbers and identification documents, a practice already observed in other Claude functionalities. Health or financial data will only be shared upon explicit user authorization during the session.
This personal agent is classified as a 'research pre-model' and is subject to continuous improvements. Anthropic assures that the service will undergo rigorous stress testing. Additionally, Claude Duet integrates the company's bug bounty program, where vulnerabilities detected by third parties will be duly considered.
