Claude Duet never sees your passwords, even when it signs in

When a site asks for a password, you type it into a form in the Claude app, which encrypts it with a single-use key that only a separate cloud browser machine can open. Anthropic says the Claude Duet model never sees the password and can’t save or repeat it.
A new classifier, run apart from the main model, checks every message Duet drafts before it goes out. Sign-in credentials, full card numbers and government ID numbers never get shared, even if you ask. Health or finance details need your consent. The safeguards “will occasionally miss something or flag a harmless message,” Anthropic says.
Trading securities and moving money between accounts stay off-limits even with your approval. Stock trades stay a job for you.
Up next
Texas sues OpenAI and Sam Altman over ChatGPT safety
“Lying to users and harming Texans” is how Texas AG Ken Paxton frames his suit against OpenAI and Sam Altman, which targets ChatGPT’s safety for teenage users. Does a teen in your house use ChatGPT freely or only with set limits?
Related stories
- Anthropic’s Claude Duet asks before it pays, sends or signs
- Anthropic test agents reached a State Department visa form
- Local MCP servers run outside Claude’s VM sandbox
- An AI agent found a hole in a gym’s booking system, then used it
- 1Password lets AI agents log in without seeing passwords
- Google opens Gemini’s agent sandboxes to Cloud customers
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.

