Safety Always-on agent by Anthropic
Is Claude Duet safe?
Short answer It depends on what you connect. Anthropic says it asks before sensitive actions, it does not see your passwords, training on your data: yes, unless you opt out; these are the vendor's own statements, not tests.
6 of 6 safety questions answered on Anthropic's own pages , Asks before sensitive actions: Yes , Sees your passwords: No , Trains on your data: Yes, unless you opt out
What Anthropic states about connecting your accounts. AideDex reports the vendor's published statements, each linked to its source; it does not test or audit them. "Not published" means we found no statement, not that the answer is no. Facts checked up to .
- Status Waitlist
- Where US (Waitlist in the US)
- Plan $20/mo (Cheapest plan that includes it: Pro)
The six questions
6 of 6 answered-
What data can Claude Duet access?
Uses the connectors you set up in your Claude account (e.g. Gmail, Google Calendar, Google Drive, Slack, QuickBooks, HubSpot, Docusign, Calendly). You start every connection and can disconnect any app at any time; memory of earlier information is kept after disconnecting. Can search your past Claude chats; turn off "Search and reference chats" in Settings > Memory to stop it. support.claude.com · checked -
Does Claude Duet ask before sending, buying or deleting?
Yessupport.claude.com · checkedDesigned to show what it will do and wait for approval before sending a message, paying, signing or deleting. Full quote
Designed to show what it will do and wait for approval before sending a message, paying, signing or deleting. Trading securities, moving money between accounts or accessing someone else’s account are off-limits even with approval. -
Can Claude Duet see my passwords?
Noclaude.com · checked"It never directly sees your passwords or sign-in tokens." Full quote
"It never directly sees your passwords or sign-in tokens." Site passwords are typed into a separate form on your device that sends them straight to the browser. -
Does Claude Duet use my data to train AI models?
Yes, unless you opt out Duet runs on your consumer Claude Pro/Max account; Anthropic lists model training as "Opt-out" on those plans. claude.com · checked -
Where does Claude Duet run, and is it isolated?
Runs on its own cloud computer (virtual machine with its own browser); approvals, logins and safety checks run there. claude.com · checked -
How long does Claude Duet keep my data?
Claude consumer chats you delete are removed from history immediately and from back-end storage within 30 days; if you allow model improvement, data may be kept de-identified for up to 5 years. No Duet-specific retention period is published. privacy.claude.com · checked
Other things Anthropic states
3 statements- A dedicated classifier rejects sharing the strictest tier of information (card numbers, government IDs); health and finance data are sent only on explicit instruction. claude.com · checked
- Penetration tested by red teams and an external firm; covered by Anthropic’s public bug bounty. claude.com · checked
- Prompt injection: Duet is built to treat what it reads (emails, web pages) as information, but Anthropic says no defense is perfect, which is one reason consequential actions need approval. support.claude.com · checked
Compare and read on
- Is ChatGPT dots safe? ChatGPT dots vs Claude Duet
- Is Meta Muse safe? Claude Duet vs Meta Muse
- Is Grok Bot safe? Claude Duet vs Grok Bot
- Is Google Labs CC safe? Claude Duet vs Google Labs CC
- Is OpenClaw safe? Claude Duet vs OpenClaw
All alternatives to Claude Duet · Is it safe to connect Gmail to an AI agent? What each vendor says it does · How facts are checked