Anthropic Releases Claude Duet Agent, but Enterprises Hesitate Over Security Risks

Anthropic has entered the personal AI agent market with the launch of Claude Duet in research preview. Built on the company’s Claude Opus 5.5 foundation model, the proactive agent is designed to execute tasks autonomously—even when the user steps away—while verifying critical actions before proceeding on their behalf.

Targeted at professionals and small business owners, Claude Duet is accessible as a standalone mobile application for iOS and Android, on the web via Claude.ai, and across Anthropic's paid Pro and Max subscription tiers.

The Proactive Agent Race

With Duet, Anthropic joins a rapidly escalating competition among leading AI developers to field autonomous personal work assistants. The rollout follows OpenAI’s September launch of its personal agent Dots and Google’s release of Gemini Agent on October 8.

The wave of recent launches follows the debut of Meta Muse, an assistant capable of managing schedules, booking travel, handling shopping, and sending emails. Like Anthropic, Meta has also been directing its agent toward small business workflows.

Emphasizing Guardrails and Restraint

Industry analysts point out that while Anthropic’s entry comes after rivals, the company has differentiated itself by embedding tighter guardrails. William McKeon-White, an analyst at Forrester Research, highlighted the system's measured release, noting that Duet’s standout quality is its restraint amid widespread concerns over unpredictable agent behavior.

Anthropic equipped Claude Duet with a dedicated classifier and safety-system prompts designed to prevent data leaks. The system automatically declines requests to expose sensitive personal records, such as government identification numbers and credit card details.

For commercial transactions, users can authorize Duet to purchase business supplies through a integrated Stripe digital wallet. However, the system strictly prohibits financial transactions involving securities trading or fund transfers between separate accounts.

Enterprise IT Remains Cautious

Despite the consumer and small business appeal, large corporate adoption faces steep hurdles. Enterprises remain skeptical about granting autonomous agents direct access to internal corporate workflows.

Paul Baier, CEO and analyst at GAI Insights, noted that large enterprise customers are not expected to deploy Duet in the near term, with some potentially avoiding it altogether. Many organizations have yet to enable Claude Cowork, a primarily non-proactive workspace agent introduced by Anthropic in January.

The core concern centers on governance and data security. While standard conversational AI chats present manageable risks, proactive agents connect directly to operational IT systems and handle proprietary data. IT security teams are stepping in to halt widespread rollouts until governance frameworks can adequately monitor automated actions across enterprise networks, meaning adoption among companies with over 1,000 employees will remain limited for the foreseeable future.