Most RIAs already own Microsoft 365. Far fewer have configured it to do the one thing the SEC now expects it to do: enforce, evidence, and retain the safeguards your written policies promise. The gap between “we have M365” and “we can demonstrate the safeguards rule to an examiner” is almost entirely a deployment problem: the right services, licensed correctly, switched on and documented.
That gap closed in 2026. The amendments to Regulation S-P adopted in May 2024 are now in force across the entire adviser population: larger entities had to comply by December 3, 2025, and smaller entities by June 3, 2026. Both dates have passed. The amendments require a written incident response program reasonably designed to detect, respond to, and recover from unauthorized access to customer information, notification to affected individuals as soon as practicable and no later than 30 days after becoming aware of the incident, oversight of service providers, and records documenting your compliance with the safeguards and disposal rules. The SEC has stated this will be an examination priority.
This article is the Microsoft 365 build. Work through it, stand the controls up yourself, and then engage MTradecraft for the verification scan described at the end, where we measure the live tenant against the CIS Microsoft 365 Foundations Benchmark and produce the evidence archive your exam file needs. The companion piece covers the same architecture in Google Workspace.
A note on settings. We work hard to reference the correct license tiers, portal locations, and configuration steps, and we are confident they are accurate and close to the mark as of publication. That said, Microsoft renames products, repackages licensing, and reorganizes the admin centers regularly, so a menu may have moved or a setting may carry a new name by the time you read this. By and large these instructions will still get you where you need to be; treat the specific click-paths as a close guide rather than a guarantee, and confirm against Microsoft’s current documentation before you rely on a given screen.
The Map: Which Rule Drives Which Control
Before touching the admin center, anchor every setting to a rule. “Best practice” is not a finding an examiner can use; a control mapped to a regulation is. Four authorities do most of the work for an adviser.
| Authority | What it requires | M365 surface |
|---|---|---|
| Reg S-P (Safeguards) | Written admin/technical/physical safeguards to protect customer information; incident response program to detect, respond, recover; breach notice as soon as practicable, no later than 30 days after awareness. | Entra ID access controls, Conditional Access, Purview DLP & labels, Defender detection, Audit. |
| Reg S-P (Disposal & Records) | Records documenting compliance with the safeguards and disposal rules; secure disposal of customer information. | Purview retention & disposition, Audit log, configuration evidence. |
| Rule 204-2 (Books & Records) | Retain electronic communications and required records for at least 5 years; first 2 years readily accessible; preserved against alteration. | Exchange Online archive, Purview retention, litigation hold. |
| Reg S-ID (Red Flags) | Identity-theft prevention program for covered accounts; detect and respond to red-flag patterns. | Identity Protection, DLP for account/SSN data, sign-in risk monitoring. |
| Rule 206(4)-7 (Compliance) | Policies reasonably designed to prevent violations; annual review. | The whole tenant: every control below is the technical leg of a written policy. |
A note on scope: Reg S-P applies broadly to covered institutions. Reg S-ID applies only where the firm is a “financial institution or creditor” that offers or maintains “covered accounts”; most RIAs that permit client-directed transfers do, but the rule requires you to make that determination periodically, not treat it as an automatic assumption.
Licensing reality. Microsoft 365 E3 (commercial-with-Teams list price roughly $36/user/mo, rising to about $39 on July 1, 2026) carries the identity and records foundation. E5 (about $57, rising to about $60) adds the detection, automation, and advanced data-protection layers the safeguards rule’s “detect and respond” language increasingly demands. (Microsoft also lists separate “no-Teams” SKUs, so confirm your exact subscription.) You do not need E5 on every seat; a common RIA pattern is E3 firm-wide with E5 (or the Purview/Defender add-on suites) on executives, IT, and anyone touching the most sensitive data. Note that Microsoft 365 E5 Compliance and E5 Security were renamed the Microsoft Purview Suite and Microsoft Defender Suite effective October 1, 2025.
Identity and Access: Who Can Get In, and on What Terms
Identity is the first safeguard an examiner tests, because credential theft is the single most common path to customer data. This is Microsoft Entra ID (the service formerly called Azure AD). Entra ID P1 is bundled in E3 and covers Conditional Access and MFA; P2 (in E5) adds risk-based policy, Privileged Identity Management, and access reviews. PIM is also available through Entra ID Governance where that is licensed instead of P2.
Multi-factor authentication, enforced for everyone
Why. Microsoft’s own telemetry shows MFA blocks the overwhelming majority of credential-based attacks. The safeguards rule expects access controls reasonably designed to prevent unauthorized access; an examiner reads “no enforced MFA” as a deficiency on its face.
How.
- Build a Conditional Access policy requiring MFA for all users, all cloud apps; do not rely on per-user MFA or leave it to security defaults alone.
- Add a second policy that blocks legacy authentication (POP, IMAP, basic auth), which silently bypasses MFA.
- For Global Admin, Exchange Admin, and other privileged roles, require phishing-resistant MFA: FIDO2 keys, passkeys, or Windows Hello, not SMS.
- Configure two break-glass accounts excluded from the policies, with long random passwords stored offline and alerted on use.
Conditional Access as the policy engine
Why. Conditional Access is where “administrative safeguards” become technical fact: it decides which user, on which device, from which location, under what risk, may reach customer data. It is the most evidence-rich control you own.
How.
- Require a compliant or hybrid-joined device (enforced via Intune, below) for access to email and SharePoint.
- Define named locations; block or step-up sign-ins from outside expected geographies if your firm is fully domestic.
- With P2, enable sign-in risk and user risk policies so a risky login is forced to re-authenticate or is blocked; this is the Reg S-ID “detect” leg made operational.
- Start every policy in report-only mode, review the impact, then enforce. Export the policy set; it is exhibit-grade evidence.
Privileged Identity Management and access reviews
Why. Standing administrative access is a finding waiting to happen. PIM makes privileged roles eligible rather than permanently active, granted just-in-time with approval and justification. Access reviews give you the recurring, documented “annual review” the compliance rule wants, for entitlements, not just policies.
How.
- Convert Global Admin, Exchange Admin, SharePoint Admin, and Security Admin to eligible assignments in PIM with time-boxed activation, approval, and MFA.
- Keep no more than two to four permanent Global Admins; everything else is just-in-time.
- Schedule quarterly access reviews of privileged roles and guest accounts; retain the completed reviews as compliance artifacts.
Data Protection: Classifying and Containing Customer Information
The safeguards rule protects “customer information.” You cannot protect what you have not classified, and you cannot prove protection without controls that act on the classification. This is Microsoft Purview. Core capabilities, manual sensitivity labels and DLP across Exchange, SharePoint, and OneDrive, are in E3. Automatic labeling, endpoint DLP, and Teams-chat DLP require E5 or the Purview Suite add-on.
Sensitivity labels
Why. Labels attach classification, and, optionally, encryption and access restrictions, to the data itself, so protection travels with a file even when it leaves the tenant. This is the difference between a policy that says data is “confidential” and a control that enforces it.
How.
- Define a short taxonomy:
Public/Internal/Confidential/Restricted. Resist more than four. - Apply encryption and an access policy to
ConfidentialandRestrictedso only authorized users can open them. - With E5, configure auto-labeling to detect SSNs, account numbers, and US PII and apply
Confidentialwithout relying on user behavior. - Publish a default label so nothing is created unclassified.
Data Loss Prevention (DLP)
Why. DLP is the technical control that stops customer information from leaving by accident or exfiltration, the exact harm the safeguards rule is written to prevent. It also produces the alert stream your incident response program needs to “detect.”
How.
- Enable DLP policies that detect SSNs, US bank/financial account numbers, and US PII across Exchange, SharePoint, and OneDrive.
- Set the action to block external sharing of matching content, with policy tips so users learn in the moment.
- Note the licensing trap: E3 DLP protects files shared through Teams (they live in SharePoint) but does not inspect Teams chat messages or endpoints; that requires E5. Document which surfaces you cover.
- Route DLP alerts to a monitored mailbox or your SIEM; the alerts are evidence the detection control is live.
Encryption — in transit and at rest
Why. Encryption keeps customer information unreadable if an email is intercepted or a device is lost. The safeguards rule expects technical protection for customer information both in transit and at rest; encryption is the most direct answer. M365 encrypts service data at rest by default and email in transit over TLS; the compliance work is in adding the layers you control.
How.
- Enable Purview Message Encryption (included in E3/E5) so users can send encrypted or “Do Not Forward” mail; apply it automatically via mail-flow rules on messages containing sensitive data.
- Use sensitivity-label encryption so
ConfidentialandRestrictedfiles carry encryption and an access policy with them wherever they travel. - For the most sensitive tenants, evaluate Customer Key (tenant-level bring-your-own-key) or Double Key Encryption (E5) where you must hold a key Microsoft cannot access.
- Confirm devices enforce disk encryption (BitLocker/FileVault) through Intune.
Archiving and Recordkeeping: Retention That Survives Rule 204-2
Two distinct obligations converge here: Rule 204-2 requires required records and electronic communications to be preserved for at least five years (the first two readily accessible) and protected against alteration; the amended Reg S-P requires records documenting your compliance with the safeguards and disposal rules. Exchange Online Plan 2 (in E3) gives every user an in-place archive; Purview retention preserves it against user deletion.
Exchange Online archiving
How.
- Enable the archive mailbox for every user and turn on auto-expanding archive, which can grow to roughly 1.5 TB (including Recoverable Items) so routine mailbox limits don’t force deletion.
- Confirm mail flow journals or retention captures both sent and received business communications, including Teams chat where used for business.
Retention policies, labels, and legal hold
Why. A retention policy preserves records against user deletion for the retention window; a user cannot delete an email out from under the rule (this is deletion-protection, not 17a-4-style WORM storage). Disposition then provides the secure, documented deletion the disposal rule expects. Together they answer both the books-and-records rule and the Reg S-P recordkeeping requirement. Core retention policies and manually published retention labels are in E3; auto-apply retention labels and disposition review require E5 (Purview Suite).
How.
- Create a Purview retention policy of at least 5 years (set to your firm’s policy; many advisers choose 6–7) covering Exchange, SharePoint, OneDrive, and Teams.
- Configure the policy to retain and then delete, so disposition is automatic and logged.
- Place litigation hold on departing employees and any custodian relevant to an inquiry; hold overrides retention deletion.
- Use eDiscovery (Standard in E3; Premium in E5) to produce records on request; retain the search and export logs.
Detection, Audit, and Email Security: The Detect-and-Respond Leg
The amended Reg S-P is explicit: your program must detect, respond to, and recover from unauthorized access. In practice, audit logging and threat detection are how most firms meet that requirement and show that they did. This is also where MTradecraft’s risk assessment and threat-detection process validation focuses during the verification scan.
Unified audit log and Purview Audit
Why. If you discover an intrusion late, which is the norm, short log retention blocks root-cause analysis and the 30-day breach-notice determination the rule now requires. Audit (Standard) retains logs for 180 days (for logs generated on or after Oct 17, 2023); Audit (Premium), in E5, extends retention to one year for Exchange, SharePoint, OneDrive, and Entra audit records (longer with add-on) and surfaces investigation-critical events.
How.
- Confirm the unified audit log is enabled (on by default in current tenants, but verify; older tenants are not).
- If you carry E5, enable Audit (Premium) for one-year retention and critical event logging on privileged accounts.
- Decide retention against your IRP: you need enough history to reconstruct an incident and make a defensible notification decision.
Microsoft Defender — detection and response
Why. Defender is the operational core of the “detect and respond” requirement. E3 carries Defender for Endpoint Plan 1 (next-gen AV, attack-surface reduction); Defender for Office 365 Plan 1: the Safe Links / Safe Attachments email layer, is being added to E3 in Microsoft’s 2026 packaging update, rolling out through mid-2026 and completing by August 1, 2026. E5 carries Defender for Endpoint Plan 2 (EDR), Defender for Office 365 Plan 2, automated investigation, Defender for Identity, and Defender for Cloud Apps for tenant-wide anomaly detection: the difference between “we hope we blocked it” and “we can show what happened.”
How.
- Confirm Defender for Endpoint is deployed and reporting on every managed device.
- Verify whether Defender for Office 365 Plan 1 (Safe Links, Safe Attachments) is active in your tenant; on E3 it arrives via the 2026 rollout (complete by Aug 1, 2026) and applies built-in protection by default; before that it required an add-on or E5. Tune the built-in protection preset and add exclusions if you run a third-party gateway.
- With E5, enable automated investigation and response and Defender for Cloud Apps anomaly policies; route alerts to your IR workflow.
Email authentication — SPF, DKIM, DMARC
Why. Business email compromise and domain spoofing are leading vectors for the credential theft and fraudulent-instruction harms the safeguards rule targets. These are free DNS controls; their absence is an avoidable finding.
How.
- Publish a correct SPF record for every sending domain.
- Enable DKIM signing in the Defender portal for each domain.
- Publish a DMARC record and move it to
p=rejectafter monitoring; enable spoof and impersonation protection in your anti-phishing policy.
Device management — Intune
Why. The “technical and physical safeguards” language reaches the endpoint. Intune is what lets Conditional Access trust a device, and what enforces encryption and lock-screen controls on the machines that hold customer data.
How.
- Enroll firm-owned devices and publish a compliance policy requiring BitLocker/FileVault, a PIN, and minimum OS patch level.
- Reference device compliance in Conditional Access so a non-compliant device cannot reach data.
- Configure remote wipe for lost or stolen devices, a documented disposal/loss control under Reg S-P.
AI Governance and Capture: Copilot, ChatGPT, and Claude as Records
Your staff are already using Copilot, ChatGPT, and Claude whether or not there is a policy for it. FINRA has confirmed (Regulatory Notice 24-09) that its rules are technology-neutral and apply to generative AI used for firm business, and its 2026 oversight report suggests keeping logs of prompts and outputs. For an adviser, AI-assisted business communications are records subject to the same recordkeeping, supervision, and safeguards expectations as any other channel. Microsoft Purview is where you make that governable.
Capture and govern Copilot with Purview (DSPM for AI)
Why. Microsoft 365 Copilot prompts and responses are captured in the unified audit log and stored in the user’s mailbox, which means they can be retained, held, searched via eDiscovery, and reviewed through Communication Compliance, exactly like email. That turns AI activity from an ungoverned blind spot into a records and supervision surface you can defend.
How.
- Use DSPM for AI to discover where AI is used and surface sensitive data exposed to it.
- Apply a retention policy for Copilot & AI interactions so prompts and responses are preserved for your records window.
- Confirm Copilot activity is searchable in eDiscovery and, where you supervise communications, add it to Communication Compliance policies.
Extend to third-party AI, and block data leakage
Why. The larger risk is staff pasting client data into public ChatGPT or Claude. Purview can extend visibility and capture to outside tools, but this is not automatic, and honest scoping matters for the exam file.
How.
- For governed capture of enterprise AI, register the enterprise editions (ChatGPT Enterprise, Claude Enterprise) with the required licensing and collection policies; consumer/unmanaged AI does not provide the same prompt/response coverage.
- Where Endpoint DLP is licensed and devices are onboarded, configure policies that warn or block users from pasting SSNs, account numbers, or other sensitive data into third-party generative-AI sites through the browser.
- Document which AI surfaces you capture and which you only block; the boundary is part of your evidence.
Where MTradecraft Comes In: Measured Against the CIS Benchmark
The framework above is what you deploy. Verification is what makes it defensible, and that is independent work, by design. Once the tenant is configured, MTradecraft performs a verification assessment that measures the live environment against the CIS Microsoft 365 Foundations Benchmark, the consensus standard examiners and auditors recognize, across its Level 1 and Level 2 profiles spanning Entra ID, Exchange Online, SharePoint, Teams, Purview, Defender, and Intune.
As part of MTradecraft’s risk assessment and threat-detection process, the verification does four things a self-attestation cannot:
- Tests every control against a named benchmark: each setting is pass/fail against a specific CIS recommendation, not an opinion.
- Captures exhibit-grade evidence: configuration exports, Graph/PowerShell output, and screenshots for each control, archived for the exam file.
- Maps each finding back to the rule: every gap is tied to Reg S-P, Rule 204-2, Reg S-ID, and Rule 206(4)-7 so remediation is prioritized by regulatory exposure, not severity alone.
- Delivers a remediation roadmap: ranked, with example remediation language your CCO can drop into the policies the controls support.
The result is evidence for your exam file that the safeguards your policies promise are actually configured and monitored, alongside the records the amended Reg S-P requires you to keep. MTradecraft is vendor-independent and 100% client-funded; we do not resell Microsoft licensing or MSP services, so the verification has no incentive but accuracy.
Stand up the architecture above, then book the CIS-benchmarked verification scan to convert your configuration into exam-ready evidence. Part II of this series covers the same build in Google Workspace.