Trust & security
An AI in your inbox should have to earn it.
Mercury reads your most commercially sensitive data: your mail, your calendar, your CRM. Here is exactly how much power it has, and where it stops.
You approve everything that leaves
Draft-don't-send and propose-don't-write are enforced product behaviors, not preferences. Mercury prepares drafts in your Outlook Drafts folder and proposes CRM changes for your click. It has no capability to email a customer or write a record on its own.
Your data stays in your region
EU and North America data residency options, so your mail and CRM context are processed where your compliance team expects them to be. Mercury is powered by Anthropic's Claude models, with EU data residency available.
Hardened against hostile email
An assistant that reads inbound mail can be attacked through inbound mail. Mercury is built with deliberate safeguards against prompt-injection emails that try to hijack it into leaking data or taking actions, and every action it takes is recorded in a full audit log.
Least privilege, human in the loop
Mercury requests only the access it needs, and sensitive threads are flagged and briefed rather than auto-drafted. The judgment calls stay with you, by design.
Have your security team grill us.
Book a demo and bring the hard questions. We built Mercury for buyers who ask them.