Why a stolen password can expose every customer chat, and how two-factor auth, lockouts and role limits protect a support dashboard on AI SNS CHAT.
A customer-support dashboard holds more than most teams realise: chat transcripts from your website, WhatsApp and Messenger, lead details, and the instructions that control what your AI agent says. If one password leaks, all of it is exposed. This article explains how two-factor authentication and login protection work in practice, what to switch on first, and how these layers are set up in AI SNS CHAT.
Why the support dashboard is a target
Attackers rarely care about chat software for its own sake. They want what sits inside it.
- Customer data. Transcripts often contain names, emails, phone numbers, order details and sometimes things customers should not have typed but did.
- Control of the agent. Someone with dashboard access could change the persona, edit the knowledge base or alter the handoff rules, so your own bot starts giving customers wrong answers.
- Connected systems. A dashboard linked to a shop, a CRM or a calendar is a stepping stone to those tools.
The usual way in is dull: a reused password from another breached service, a phishing email, or a shared login that a former colleague still knows. None of this needs clever hacking. That is why the first layer to harden is the login itself.
How two-factor authentication works
Two-factor authentication (2FA) asks for two different kinds of proof. The first is something you know, your password. The second is something you have, usually your phone.
The most common form is a TOTP code (time-based one-time password). An authenticator app on your phone shows a six-digit code that changes every 30 seconds or so. The app and the server share a secret set up once, usually by scanning a QR code, and both calculate the same code from that secret and the current time.
What this buys you:
- A stolen or guessed password alone is no longer enough.
- The code works without mobile signal, because it is calculated on the device.
- Codes expire quickly, so one captured over someone's shoulder is useless a minute later.
What 2FA does not do
Be realistic about the limits. 2FA does not help if someone steals an already-logged-in session, or if the user is tricked into typing a code into a fake login page in real time. It reduces risk sharply, but it sits alongside other controls rather than replacing them.
The weak links 2FA does not fix
A strong login still leaves gaps in how teams actually work. Check these:
- Shared accounts. One login used by four agents means nobody can tell who changed what, and removing one person means changing the password for everyone.
- Over-broad access. A part-time moderator does not need to edit billing or delete agents.
- Forgotten keys. API keys pasted into old scripts can outlive the employee who created them.
- Unattended guessing. Without limits, a bot can try thousands of passwords against a login page.
- Public chat links. An agent link intended for staff but shared widely is its own exposure.
Each one has a matching control, covered below.
Lockouts and rate limits against guessing
Password guessing is cheap to automate. The defence is to make it slow and loud.
An account lockout stops guessing on a single account: after a set number of failed attempts, the account is locked for a period. A rate limit slows guessing at the network level by capping how many attempts one address can make in a window.
Here is a hypothetical to show why it matters. Suppose a bot can try one password per second against an account with no protection. That is 3,600 guesses an hour, all day. If the account locks after 10 failures for 30 minutes, the same bot gets 10 guesses, then waits. Over a full day that is a few hundred guesses at most instead of tens of thousands, and every one of them still has to get past 2FA.
- 01Password entered
- 02Lockout and rate limits
- 03Second factor (TOTP code)
- 04Role-based access
Limit what each login can do
Even with perfect authentication, someone will eventually make a mistake. Role-based access keeps that mistake small.
The principle is least privilege: give each person the smallest set of permissions that lets them do the job.
| Person | What they need | What they do not need |
|---|---|---|
| Support agent | Read chats, take over, reply | Edit billing, delete agents |
| Reviewer or manager | View chat logs and analytics | Change the persona or integrations |
| Workspace owner | Everything | Nothing is excluded, so keep the group small |
Pair roles with an activity log that records who created, edited, deleted or uploaded something and when. When a knowledge-base page changes unexpectedly, you can find out who did it instead of guessing.
Protect the chat side as well as the dashboard
Login protection covers the people who run the system. Your customers' side needs attention too.
- Prompt-injection and abuse filters. Someone chatting with your agent may try to talk it out of its instructions. Filters on the agent screen the messages before they reach the model.
- Per-device rate limits. One visitor flooding your widget should not drain your credits or crowd out real customers.
- Gated links. If you share an agent link internally, protect it with a password or per-user sign-in instead of leaving it open.
- Scrubbed knowledge. Passages shown to strangers should not leak personal data that slipped into your documents.
A practical rollout checklist
You can do most of this in an afternoon.
- Turn on 2FA for every owner and admin first. These accounts can do the most damage.
- Then extend it to the whole team. Make it a condition of getting access, not an optional extra.
- Retire shared logins. One person, one account. Remove accounts the day someone leaves.
- Review roles. Move anyone who only reads chats to a read-oriented role.
- Audit API keys. Revoke any key you cannot attach to a current purpose.
- Check your agent links. Anything meant for staff should be gated.
- Test recovery. Make sure two people can regain access if a phone is lost.
- Schedule a quarterly review of members, roles and keys, ten minutes on the calendar.
Doing this in AI SNS CHAT
AI SNS CHAT covers each layer above. Here is what it provides and where the limits are.
Sign-in and two-factor. You can sign in with email, Google or a phone SMS code. Users can add TOTP two-factor authentication with a standard authenticator app, and the platform can require administrators to use it. If you cannot find the screen, ask Synthia, the built-in fleet manager. She can give a guided screen tour that navigates to the right page and highlights the control, by voice or text.
Lockout. By default, 10 failed logins lock an account for 30 minutes. The platform also publishes rate limits for its API, for example 20 authentication requests per 5 minutes per IP address. Passwords are stored with bcrypt and stored secrets are encrypted with AES-256-GCM.
Roles and accountability. Workspaces have Owner, Co-owner, Member, Support and Viewer roles, with per-member permission overrides. Give your chat reviewers Viewer or Support and keep Owner for as few people as possible. Custom roles and SSO are Corporate-plan items and are hidden at present, so do not plan around SAML sign-in. Owners and admins can open Workspace settings → Activity to see every create, edit, delete or upload with the person's name, email, feature and time. Each data source also shows who added it and who last edited it.
API keys. Keys are created under Workspace settings → API keys on Hobby or higher. A key is shown once and can be revoked at any time. Revoke old ones during your quarterly review.
Customer-facing protection. On each agent's security page you can turn on prompt-injection and abuse filters and per-device rate limits. Shareable agent links can require a password or per-user sign-in, and can be restricted to your office network. Platform-level anti-fraud controls, such as blocking disposable email addresses and suspicious IPs, are managed by the platform.
Honest scope. Practices are designed with reference to GDPR and APPI. AI SNS CHAT is not SOC 2, ISO 27001 or HIPAA certified, and the site says so. Read the details on the security page, compare what each plan includes on pricing, or browse everything on the features page. To try it with a small team, you can sign up.
Frequently asked questions
Which kind of two-factor authentication does AI SNS CHAT support?
It supports TOTP, the six-digit codes produced by standard authenticator apps. Administrators can be required to use it. Check your own account to confirm it is enabled for every owner and admin.
What happens after too many wrong passwords?
By default, 10 failed logins lock the account for 30 minutes. This slows automated guessing considerably. Keep a second owner on the workspace so a lockout never leaves you without an administrator.
Can I sign in to AI SNS CHAT with GitHub or single sign-on?
No. Sign-in options are email, Google and phone SMS. SSO is listed as a Corporate-plan item but is hidden, so do not rely on it for your rollout.
How can I see who changed something in my workspace?
Owners and admins can open Workspace settings → Activity. It records every create, edit, delete or upload with the person's name, email, feature and time.
Does two-factor authentication protect my agent from people abusing the chat?
No, those are separate layers. 2FA protects the dashboard login. Customer-side abuse is handled by each agent's prompt-injection and abuse filters and per-device rate limits.


