Last updated: 9 November 2026
Not without limits. Never give an AI agent a password vault or full mailbox access. Start with a single folder or an approved tool with no account access, and put in place a short policy, MFA with restricted app consent, and audit alerts so you can see what the agent does and switch it off.
What happened
Alex Stamos, Meta’s former security chief, told the Big Technology Podcast it is “totally insane” to give an AI agent things like bank logins, as reported by Business Insider. This followed the launch of Meta’s personal AI agent, Muse. Some users are reportedly giving it their password vaults and whole inboxes. Meta says Muse runs on an isolated virtual machine and advises users to review activity and limit access. Several issues are described as reported and not independently verified by ABCOM.
Safe versus risky
| What you share | Risk | Why |
|---|---|---|
| A password vault | Critical | Access to every account in it, with no way to limit scope |
| Full mailbox access | High | Years of private and client information become readable |
| A single shared folder | Medium | Limited scope, but still check what is inside |
| An approved tool with no account access | Low | It only sees what you paste in, under your policy |
Five steps to take this week
Write a policy
A one-page AI acceptable use policy naming approved tools, banned data types and who to ask.
Review app consent
In Entra, check Enterprise applications, remove unknown apps and restrict user consent so new apps need admin approval.
Enforce MFA
Require MFA for everyone and block sign-ins from unmanaged devices where you can.
Log and alert
Turn on audit logging and alert on new app consents, mailbox rules and unusual sign-ins.
Also ban password vault sharing with AI tools, plainly, in the policy and in staff training. Our guide to an AI acceptable use policy explains what to include.
How ABCOM helps
We can check your Microsoft 365 app consent, MFA and logging and help you draft an AI policy. The source also recommends keeping cloud AI as your primary but adding a tested, independent failover, and notes that open source is not automatically safer: if you self-host, you own the updates, access control and monitoring. Our managed IT services can run these controls for you.
Quick answers
Is it safe to give an AI agent access to our email?
Full mailbox access is a high risk. Start with a single shared folder or an approved tool with no account access, and review what the agent does.
Should we stop using cloud AI?
No. Cloud AI is a sound primary choice. Add a tested, independent failover so you are not reliant on one provider.
What is an AI agent?
Software that takes actions for you, such as reading email, filling in forms or booking services, rather than only answering questions.
Unsure what AI tools your staff already use? Talk to ABCOM about a free review.
Related answers and guides
- Giving an AI agent your logins (ABCOM)
- Do we need an AI acceptable use policy, and what should it say?
- ABCOM managed IT services
- Vulnerability assessment
- All ABCOM FAQs
Sources and further reading
Related questions
Want this sorted for you?
Talk to our Burgess Hill team about our AI Readiness Sprint. Free, friendly and no obligation.