Anthropic Launches Claude Duet With Approval-Gated Financial Actions

Anthropic introduced Claude Duet, a research preview of an always-on AI assistant designed to handle ongoing tasks—including shopping and limited financial actions—but only with explicit, approval-gated permission at each sensitive step.

Key points:

• Claude Duet can take actions like completing a purchase through Stripe's Link or Shopify's Shop Pay, but it is explicitly blocked from independently trading securities or transferring money between accounts.

• Every sensitive action requires the user's direct approval before it executes—there is no standing, unsupervised authority to spend or move money.

• Anthropic paired the launch with a new sensitive-information classifier, a system that automatically detects when a conversation touches financial, health, or other high-stakes information and applies extra caution—like a filter that recognizes when a conversation has gotten sensitive and tightens the rules automatically.

This is a direct answer to the trust problem facing "always-on" AI assistants. Anthropic is betting that permission-gated, narrow autonomy will earn more user and enterprise trust than broad, ad-supported autonomy—a pointed contrast to Meta's Muse companion approach.

The design also positions Anthropic competitively against OpenAI's more research-stage, less consumer-facing agent work this week.

If you're curious about always-on AI assistants but wary of handing over financial access, Claude Duet's approval-gated design is worth testing specifically because every sensitive step still requires your sign-off.

Why It Matters: By requiring explicit approval for every sensitive action, Anthropic is attempting to solve the trust problem that could otherwise block mainstream adoption of always-on AI assistants.