Office 365 incident report

The regulator wants the report in 72 hours. How much of it can you write now?

An incident report has three sections nobody can write from memory: what happened and when, how far the account could reach, and what was actually done about it. 1Security produces all three from data it already holds - up to three years of attributed activity, the live permission graph, and a log of every action taken - so the impact section is a count, not an estimate, and the report is done before the deadline is close.

The problem

The clock starts before the investigation ends.

Incident reports fail in the same three places. The timeline has gaps because the entry point was in month seven and nobody kept the history that far back. The impact section says "potentially affected" because nobody can enumerate what the account could reach - in a typical tenant that is hundreds of thousands of files through groups, links and inheritance. And the remediation section says "access was revoked" with nothing attached to prove it.

Meanwhile the deadlines are statutory. GDPR gives you 72 hours from awareness to notify the supervisory authority. NIS2 wants an early warning within 24 hours, a notification with an initial assessment within 72, and a final report within a month - each with severity, impact and indicators of compromise spelled out. Every one of those needs facts you either have on hand or spend the deadline reconstructing.

The usual result is a report written from four exports and a memory of who did what, finished at 2 a.m., with an impact section that ends in a question mark. 1Security keeps the incident as one object instead: the detection, the activity behind it, the containment actions and the review history - and the report is generated from that object, with the deadlines computed from the moment you became aware.

In practice

From the detection to the document, on one screen.

How the report comes together in 1Security, in the order a regulator reads it.

  1. 01

    Open the detection, read what happened

    Open the episode on Anomalies. The drawer leads with the detector that fired and the entity it fired on - a user, an app, an AI agent or a place - the baseline against the observed activity, the peak, and a severity earned from the score and the other open detections on the same entity, not stamped by a rule. The Activity tab holds the events in the incident window, each attributed to actor, resource, application, device and location, even if the first sign-in was ten months ago.

  2. 02

    Name the indicators and the reach

    The addresses, networks, countries, devices and client apps behind that activity become the indicators of compromise, aggregated from the same rows. The travel trail says whether the hops were plausible, improbable or impossible for one person. Open the account in Users and the permission graph resolves everything it could reach through grants, links, nested groups and inheritance - "potentially affected" becomes "could reach 214,000 files, 3,100 of them with personal data".

  3. 03

    Contain it and write it down as you go

    Respond offers the curated containment actions for the entity - block sign-in, revoke grants, quarantine the agent - and runs them through the Action Center, which keeps who requested what, when it ran, whether Microsoft confirmed it and the revert that undoes it. Every status change, severity verdict and requested action lands on the History & Comments tab on its own; your comments fill in what you checked and why.

  4. 04

    Generate the report from the drawer

    Click Report. Pick a template - the Incident record, the two-page Incident briefing, or the NIS2 Article 23 early warning, incident notification or final report - a language and a reporting stage, tick and order the sections, and write what no system can know: description, impact, root cause, mitigation, and the assessment answers printed as Yes, No or Not established. Download PDF, PowerPoint, HTML, CSV or JSON while the preview updates beside you. The text is saved per detection, so the final report a month later starts from the notification's words.

  5. 05

    Let the next report judge the deadlines

    Generating a report is itself a History & Comments event - who, when, which template and stage, which format. A staged NIS2 report reads those entries back into its Reporting obligation section: an early warning sent before the 24-hour mark says On time, one after it says After the deadline, and each Article 23(4) point is mapped to the section that answers it. A SOC platform pulls the same document over the REST API or MCP by the detection id.

Anomalies and detections: Spikes in sharing, downloads, sign-ins, consent and agent activity, scored against each tenant’s own pattern, with curated response actions.

What makes it work

The detection, the memory and the document.

The report is only as good as the detection behind it, the history underneath it and the engine that renders it.

  • Anomaly detection

    Episodes with a baseline, an earned severity, correlation across detectors, Respond actions and a History & Comments trail - the object the report is generated from.

    See the feature →
  • Reports

    One report engine for every document: templates, your branding, PDF, PowerPoint, HTML, CSV and JSON in seven languages, served over the REST API and MCP.

    See the feature →
  • Three years of audit history

    Go from 180 days of audit history to three years with 1Security - searchable, attributed, on the licenses you already have, with the raw record one click away.

    See the feature →

FAQ

Common questions.

Does this cover the GDPR and NIS2 reporting deadlines?

NIS2 Article 23 comes as three ready-made templates - the early warning, the incident notification and the final report - with the 24-hour, 72-hour and one-month deadlines computed from when you became aware and each marked as met or missed. A GDPR notification is written from the Incident record, which carries the facts Article 33 asks for: what happened, who was affected, the indicators and the measures taken. Deciding whether an incident is reportable, and to whom, stays with your counsel; every assessment you have not made prints as Not established rather than a guess.

What if the incident started months ago?

That is what the retention is for. Up to three years of history on a standard license means an entry point from last quarter is still in the record, attributed to its user, device and location - the 180-day question stops mattering.

Can we prove the remediation actually happened?

Yes. Every action taken through 1Security lands in the Action Center with the resource, the action, who requested it, when it ran and whether Microsoft confirmed it, with a revert path. The report's Response and mitigation section lists those actions next to the measures you describe in words, and the History & Comments tab shows who requested each one from the drawer and when. Actions are real Microsoft-side changes, so the export matches what the auditor sees in the admin centers.

Do we need E5 or a SIEM to get this?

No. Retention, attribution, blast radius, the detection drawer and the report run on standard Microsoft 365 licenses with a read-only connection and nothing to install. Write actions are a separate opt-in module; until you enable it, the report still has the timeline, the indicators and the reach - only the mitigation section is filled by hand.

Can our SOC pull the document instead of logging in?

Yes. The REST API renders any incident template for one detection by its id, and the MCP tool does the same for an AI assistant in your SOC platform - in PDF, PowerPoint, HTML, CSV or JSON and in any of the seven report languages. A read-only API key is enough, and every render is recorded on the detection's History & Comments tab with the key's name as the actor.

The 1Security platform

Incident Reporting is just the beginning

Incident Reporting is one screen of the 1Security platform for Microsoft 365. Locations, devices, apps, AI agents, logs, permissions, sensitive data, compliance, reports and actions run on the same read-only connection, on the licence you already have.

Write the next incident report from the detection, not from memory.

Connect read-only and the history starts building the same day. When the next incident lands, the drawer holds what happened, the Report button turns it into the notification document, and the deadlines are judged by the record - while the 72 hours are still comfortable.

Or keep writing impact sections that end in a question mark.