How to protect your Intercom account?

Intercom is no longer just a chat widget or "CRM." In 2026 it is a full customer service platform: Fin, its AI agent, answers customers directly, Copilot drafts replies for human agents, and an MCP server lets outside AI tools read conversations, contacts and tickets.

That makes Intercom one of the highest-risk SaaS apps for sensitive data. Every conversation is unstructured text written by customers, who don't know or care what your compliance team wants. They paste full card numbers to fix a failed payment. They upload a driver's license to verify an account. Patients describe symptoms. Developers share API keys while debugging.

All of it lands in your Intercom workspace, often stays there indefinitely unless someone deletes it, and is now readable by more people and more AI agents than ever. This guide covers how to lock down access with Intercom's built-in settings, where those settings stop, and how to protect the data itself with DLP and DSPM.

Why Intercom data is at risk in 2026

Breaches are more expensive and more often start with people and partners. According to IBM's 2026 Cost of a Data Breach Report, the global average breach now costs $4.99 million, a record and up 12% year over year; in the US it is $11.5 million, and healthcare stays the costliest industry at $6.64 million (Baker Donelson summary). Customer PII showed up in 52% of breaches (Passwork summary). Verizon's 2026 DBIR found the human element in 62% of breaches and a third party in 48% (Abnormal summary).

For Intercom specifically, the risk comes from five places:

• Customers overshare. They paste card numbers, SSNs, passwords and health details into chat to speed things up.
• Attachments and screenshots. ID photos, insurance cards, bank statements and bug screenshots get uploaded, and the data inside images and PDFs can't be found with text search.
• Broad teammate access. Support, success, sales and contractors often share one workspace, so one stolen login exposes every conversation that person can read.
• AI tools. Fin and Copilot work from your conversations and help content, and outside AI tools like Claude, ChatGPT and Cursor can read Intercom through the Intercom MCP server. Whatever sits in those conversations goes to the AI too.
• Integrations and exports. CRM syncs, webhooks, data exports and analytics tools copy conversation data out of Intercom, so the same sensitive data ends up in more places.

Regulators expect you to control this data wherever it lives. PCI DSS forbids storing full card numbers where they aren't needed. HIPAA applies the moment a covered entity's patient writes into chat. GDPR and the growing list of US state privacy laws give customers the right to have their data deleted, which you can't honor if you don't know where it is.

✨ How to secure your Intercom workspace with native controls

These settings are free, quick, and the foundation for everything else. Do them first.

1. Enforce SSO or two-factor authentication

Stolen credentials remain a common way into SaaS apps. Intercom supports two-factor authentication, Google sign-in and SAML SSO. Enforce SAML SSO through your identity provider (Okta, Entra ID, Google Workspace) so access is removed automatically when someone leaves. If SSO isn't an option, require 2FA for every teammate, not just admins.

2. Secure the Messenger with JWTs (Identity Verification is deprecated)

Many older Intercom security guides still tell you to turn on Identity Verification. That advice is out of date. Intercom has deprecated Identity Verification, the HMAC "user hash" method, and now recommends Messenger security with JSON Web Tokens.

Your server signs a JWT with the user's user_id and your Messenger API secret, and Intercom checks it on every request. This stops one logged-in user from impersonating another and reading their conversations. Set it up under Messenger security settings, enforce it on web, iOS and Android, and rotate the secret under Workspace > Security > Messenger (Intercom developer docs). Already on HMAC? Follow Intercom's migration guide.

3. Apply least privilege to teammates

Give each teammate the smallest role that lets them do their job. Use custom roles to restrict exports, app installs, billing and workspace settings to a few admins. Then limit which conversations each team can see, so a billing contractor can't browse VIP or healthcare inboxes.

4. Review teammate activity logs

Teammate activity logs show logins, invites, permission changes, setting changes and data exports. Review them at least monthly, and alert on bulk exports, new admin invites and logins from unusual locations.

5. Audit connected apps, AI agents and API tokens

Every App Store app, webhook, OAuth app and access token is another path out of Intercom. Remove what nobody uses, and check that the rest only have the scopes they need. Review which content sources Fin can draw on, and inventory which external AI tools connect over the Intercom MCP server.

6. Redact and delete what you don't need

Teammates can manually redact individual messages that contain sensitive data, and you can delete user data to meet GDPR and CCPA requests. This works for one-off cases. It does not scale to thousands of conversations a day, which is where the next section comes in.

Where Intercom's native controls stop

Native controls answer who can open a conversation. They don't answer what is inside it. That leaves five gaps:

1. No content inspection. Nothing flags or masks a card number, SSN or API key the moment a customer pastes it. Every teammate with access to that inbox sees it in full.
2. Attachments are a blind spot. A photo of a driver's license or an insurance card looks like any other image. Finding the data inside needs OCR.
3. Historical data piles up. Years of conversations hold sensitive data nobody has reviewed. Access controls protect it going forward but don't tell you it exists.
4. AI agents widen the audience. Once Fin, Copilot or an outside AI tool pulls up a conversation, the AI can read everything in it. Role settings don't hide sensitive details from the AI.
5. Right-to-delete is manual. Honoring a GDPR or CCPA request means finding every place a customer's data appears, across conversations, notes, attachments and the tools Intercom syncs to.

Closing these gaps takes data-level controls: DLP that inspects and redacts content in real time, and DSPM that finds and remediates what is already stored.

✨ How Strac protects Intercom data: DLP, DSPM and MCP in one platform

Strac Intercom DLP protects the data itself, at three layers: as it enters Intercom, as it sits in Intercom, and as AI agents read it. It's published on Intercom's App Store, needs no software installed, and connects in under 10 minutes.

Layer 1: Real-time DLP on every conversation

Strac inspects every message and attachment as it arrives, using its catalog of sensitive data elements: SSNs, driver's licenses, passports, card numbers, bank accounts, dates of birth, PHI, API keys and custom data types you define. OCR reads the text inside screenshots, ID photos and image-based PDFs.

You choose the mode per policy:

• Detect only: Strac flags sensitive conversations and attachments, shows findings in the Strac dashboard and sends alerts.
• Redact: Strac masks the sensitive content inside Intercom. Authorized users can still view the original in Strac's encrypted vault, and every access is logged.

A customer who pastes a card number to fix a payment gets help; the card number never sits in your inbox. That is the core control for PCI DSS PAN masking.

Layer 2: DSPM for historical conversations

Real-time DLP stops new exposure. Strac DSPM scans existing conversations and attachments, classifies what it finds, and lets you redact or delete in bulk. You get a clear picture of how much PII, PCI and PHI sits in Intercom today, and a practical way to honor deletion requests.

Layer 3: MCP DLP for AI agents

When Claude, ChatGPT, Cursor or a custom agent reads Intercom through the MCP server, Strac sits between the AI tool and Intercom. It checks every request and redacts, masks or blocks sensitive data before the AI tool sees it. Risky actions, like reading many conversations at once or sending a message, can require approval. Every request is logged with who made it, which AI tool, what data was found and what Strac did. The Intercom MCP Server guide covers the setup in detail.

Audit-ready compliance

Every finding, redaction and vault access is logged and can stream to your SIEM (Splunk, Microsoft Sentinel, Datadog). The same evidence supports PCI DSS, HIPAA, SOC 2, ISO 27001, GDPR and CCPA audits.

Beyond Intercom

Support data rarely stays in one app. The same Strac policies cover Zendesk, Slack, Gmail, Google Drive, Salesforce, HubSpot, ChatGPT and endpoints, so a card number blocked in Intercom is also caught when someone copies it into Slack or an AI chat.

Intercom security checklist for 2026

Access

• Enforce SAML SSO, or require 2FA for every teammate
• Move the Messenger from Identity Verification (HMAC) to JWT-based Messenger security on web, iOS and Android
• Rotate the Messenger API secret and store it server-side only
• Assign least-privilege roles; restrict exports and app installs to admins
• Limit conversation access by team, inbox or brand

Monitoring

• Review teammate activity logs monthly; alert on bulk exports and new admins
• Remove unused App Store apps, webhooks and API tokens
• Inventory Fin's content sources and every AI tool connected over MCP

Data protection

• Turn on real-time DLP to detect and redact PII, PCI, PHI and secrets in messages
• Enable OCR scanning for attachments and screenshots
• Run a DSPM scan of historical conversations and remediate findings
• Use MCP DLP to protect data that outside AI tools read from Intercom
• Stream DLP and audit events to your SIEM
• Document a repeatable right-to-delete process

The bottom line

Lock down access with Intercom's built-in settings, then protect the data itself. Intercom holds some of the most sensitive data your customers ever share, and in 2026 more people and more AI tools can read it than ever. Strac gives you real-time redaction, cleanup of old conversations and control over AI tools for Intercom, in one platform with nothing to install and setup in under 10 minutes.

Book a demo to see Strac redact a live Intercom conversation.

🌶️ Spicy FAQs: Intercom security

Is Intercom secure?

Intercom provides strong platform security, including SSO, 2FA, role-based access and audit logs. But security of the platform is not the same as security of your data inside it. Intercom can't stop customers from pasting card numbers or uploading ID photos; that is your responsibility under the shared responsibility model.

Is Intercom Identity Verification still supported?

Identity Verification (the HMAC user hash) is deprecated. Intercom now recommends Messenger security with JSON Web Tokens and offers a migration guide. New installations should use JWTs.

How do I redact PII in Intercom automatically?

Intercom lets teammates redact messages by hand. To redact automatically, use a DLP tool such as Strac Intercom DLP, which detects and masks PII, PCI, PHI and secrets in messages and attachments in real time.

Can customers share credit card numbers in Intercom chat?

They will, whether you allow it or not. Storing full card numbers in a support tool expands your PCI DSS scope. Real-time redaction masks the number on arrival, keeping it out of the inbox.

Is Intercom HIPAA compliant?

Intercom supports HIPAA use on certain plans with a signed BAA. A BAA covers Intercom's obligations, not what your teammates and AI tools do with PHI. Redacting PHI from conversations and attachments reduces exposure.

Does Fin or an MCP-connected AI agent see sensitive data in conversations?

AI tools work from the conversations and content they can access. If a conversation holds a card number or PHI, an AI tool that reads it sees that data too. Redacting at the source, and using Strac MCP DLP for outside AI tools, keeps it out.

How do I handle GDPR or CCPA deletion requests in Intercom?

You can delete a user's data in Intercom, but first you need to know everywhere it appears, including attachments and synced tools. A DSPM scan maps where personal data lives, so deletion is complete rather than best-effort.