Anthropic launches Claude Duet personal agent with strict security limits

The rise of personal agents

Anthropic has officially entered the personal AI agent market with the release of Claude Duet. This new tool is currently available in a research preview for professionals and small business owners. It represents a shift toward proactive technology that can complete tasks even when a user is not actively interacting with the interface.

The release follows a broader trend in the industry where major model makers are racing to provide autonomous assistants. OpenAI launched its personal agent, Dots, in September, and Google followed with Gemini Agent on October 8. These tools aim to handle daily schedules, shopping, and communication tasks for individual users.

Meta was an early mover in this space with the release of Muse. That social media giant's agent is designed to manage emails, book travel, and handle various daily chores. Anthropic is now positioning Duet to compete in this growing landscape of personal digital assistants.

How Duet handles autonomy

Claude Duet is built on the Claude Opus 5.5 model and is designed to function as a proactive agent. Unlike standard chatbots, Duet can carry out specific tasks on behalf of the user. However, the system is designed to check in with the user before executing actions to ensure it remains under human supervision.

The application is accessible through several different platforms to reach a wide audience. Users can find it as a standalone app for iOS and Android devices, or access it via the web at Claude.ai. It is also included as a feature for those subscribed to the Pro and Max plans offered by the company.

While the agent can perform complex actions, Anthropic has placed specific boundaries on its capabilities. For example, users can direct the agent to purchase items for their businesses using a digital wallet from Stripe. Conversely, the agent is strictly prohibited from trading securities or moving money between different accounts.

Safety and information guards

Anthropic is taking a more cautious approach to the release of Duet compared to some of its competitors. The company integrated a dedicated classifier and specific safety-system prompts into the tool. These features are designed to prevent the leakage of sensitive information during the agent's autonomous workflows.

The security system is specifically tuned to recognize and reject requests involving private data. This includes blocking the agent from sharing things like government IDs or credit card numbers. By filtering these requests, the company hopes to provide a safer environment for small business owners who may be handling private client data.

Industry analysts have noted that this restraint is a notable characteristic of the Duet launch. William McKeon-White of Forrester Research noted that the internet is full of examples of unintended and rogue behavior from autonomous agents. He suggested that a cautious approach is a better way to introduce these types of tools to the general public.

Enterprise hesitation and risks

Despite the new release, large-scale enterprise adoption of personal AI agents appears to be lagging significantly. Paul Baier, CEO and analyst at GAI Insights, suggests that B2B customers may not roll out Duet for months or even years. Many large companies have not even fully adopted Anthropic's previous non-proactive agent, Claude Cowork, which launched in January.

The primary hurdle for large organizations is the complexity of allowing agents to connect to internal IT systems. While letting employees use standard AI chats is considered low risk, proactive agents can automate processes involving confidential information. This creates a significant hurdle for IT and governance groups who must oversee how data flows through these systems.

Baier noted that companies with more than 1,000 people are not currently clamoring to adopt these agents in the near term. The concern lies in the potential for employees to begin prototyping apps that tap into existing infrastructure. Security teams are currently prioritizing caution over the speed of deployment for these autonomous tools.

The path forward for agents

The launch of Claude Duet signals a transition toward a world where AI can act as a representative for human users. While the immediate impact may be felt most by small business owners and individual professionals, the underlying technology is moving toward deeper integration. The goal is to move from