AAI AtlasEXCHANGE ANYWHEREAbout
AI ATLAS · LEADERSHIP LEARNING GUIDE

Enterprise Cybersecurity & Resilience

A practical guide to protecting people, platforms, data and AI.

For enterprise architects, senior leaders, product leaders and business stakeholders: understand where exposure sits, which questions to ask and what evidence supports a decision.

Curated by Ravikumar Sathyamurthy · Refreshed 13 November 2026

Start with a critical business service. Select a topic below, then use its questions with the accountable team. Priorities depend on your organisation’s risks, obligations and operating model.

Ten leadership priorities connected around trusted business operations.
The reference map connects business risk with operational responsibilities. Explore each topic below. View full-size diagram

Security at a Glance

Ten connected lenses. One business objective: reduce harm and sustain trusted operations.

Select any branch to explore guidance directly beneath it. These learning topics are an Atlas grouping, not a prescribed standard or a ranked checklist.

Turn Awareness into a Useful Conversation

1 · Establish the Consequence

Which service, people or data could be affected? Explain the business impact before discussing a tool or score.

2 · Ask for Evidence

Request a recent test, coverage result or exception record. A configured control is a starting point; test whether it works.

3 · Agree the Decision

Name the owner, action, funding and review date. Balance risk reduction with continuity, accessibility and productive work.

What Deserves Continued Attention?

Use these enduring watchpoints in reviews. This guide is not a live threat feed.

Stolen identities, persistent sessions and convincing impersonation

Review access beyond the sign-in screen, privileged permissions and verification of sensitive requests. Use an independent trusted channel for suspected deepfake impersonation; synthetic media also has legitimate uses. Include contractors, suppliers and workload identities. Test the actual revocation window rather than assuming immediate logout everywhere.

Exposed vulnerabilities and rapidly changing dependencies

Track assets, vendor advisories and known exploitation. A zero-day may have no patch: assess mitigations, exposure reduction or temporary restrictions. Verify fixes and record the impact of deferring them.

AI expands existing data and action boundaries

Review oversharing, connector scope, tool access and cross-user separation. Malicious content can reach an agent through otherwise legitimate documents or tools; validate technical enforcement beyond the model’s instructions.

Internal misuse, accidental disclosure and unsanctioned services

Review insider-risk signals with context and privacy safeguards. Discover shadow IT and AI, including personal accounts, extensions and excessive connector grants. Provide approved alternatives and address the business need alongside the exposure.

Supplier concentration and recovery dependencies

Review the services, identities and integrations on which recovery depends. Consider supplier disruption and compromise alongside internal incidents. Practise restoration and decision-making.

Protect Data According to its Purpose and Sensitivity

Identify the information → classify it → apply handling controls → test coverage → review changes.

PII, sensitivity labels and protection: what is the difference?

Personal information / PII: information that identifies a person directly or in combination with other information. Applicable legal definitions vary. Its risk depends on context, use and exposure; other sensitive data includes trade secrets and commercial plans.

Classification: your organisation defines categories and handling rules. An illustrative scheme is Public → Internal → Confidential → Restricted. Sensitivity labels implement the chosen taxonomy in supported systems. Names such as Sensitive or Protected may be used under a specific policy or government scheme; they are not interchangeable universal levels.

Protection: translate classification into least-privilege access, approved recipients, encryption where appropriate, sharing and export limits, DLP, data minimisation and approved AI destinations. Set retention and deletion separately, respecting legal holds and obligations.

Evidence: test detection accuracy, authorised access and blocked unauthorised sharing. Automated classifiers can miss information or mislabel it. Check unsupported formats, copied content and downstream systems; use synthetic data in learning exercises.

Official sensitivity-label guidance · Official sensitive-information-type guidance

How Frameworks and Tools Fit

Frameworks

NIST CSF helps organise risk outcomes. Use it to connect governance, identification, protection, detection, response and recovery.

Reference Architecture

Microsoft MCRA illustrates security architecture using Zero Trust. Apply its vendor context; it is not a universal product requirement.

Protocols & Products

Integration protocols and security tools implement parts of a design. They need ownership, appropriate scope and evidence of operation.

Explore a small set of security capability examples

Illustrative resources, not equivalent alternatives, endorsements or a shopping list. Use existing investments where adequate; validate licensing, tier, integration, coverage and release status.

CapabilityExamples with official documentationLeadership check
Threat detection & responseDefender XDR · GuardDutyWhich assets and signals are actually covered?
Cloud posture & workload protectionDefender for Cloud · Security Hub CSPM · Security Command CenterWho resolves findings and verifies improvement?
Vulnerability assessmentAmazon InspectorWhich workload types and dependencies remain outside coverage?
Security monitoring & investigationMicrosoft SentinelAre useful detections connected to a staffed response process?
Data protection & governanceMicrosoft PurviewDo the specific controls work for the data locations we use?
Insider-risk investigationMicrosoft Purview Insider Risk ManagementAre investigations proportionate, access-controlled and reviewed by authorised people?
Shadow IT visibilityMicrosoft Defender for Cloud AppsWhat remains outside discovery, and how are approved alternatives provided?

A Short Learning Route

Read the NIST overview → review one critical service using the map → follow the relevant architecture or vendor documentation → rehearse a synthetic incident → record the next action and owner.

Continue with Responsible AI for principles and accountability, or the AI Adoption Map for delivery choices. The Digital Workplace Ecosystem provides the wider service context.

Keep the Guide Useful over Time

Use the same questions in service and risk reviews. Record scope, evidence date, accountable owner and the next action; avoid an unsupported overall security score.

A practical review cadence

Routine reviews: examine critical exposure, access failures, sensitive-data handling, monitoring gaps and overdue exceptions at a frequency set by risk. Quarterly leadership review is an illustrative starting point, not a mandated interval.

Event-driven review: reassess after a material architecture or supplier change, new AI authority, an incident, an actively exploited applicable vulnerability or a relevant regulatory change. Urgent threats should not wait for a calendar review.

Resource maintenance: verify official links, release status and coverage; date material revisions and keep a short change record. This guide is manually curated, with no automatic monitoring or live threat feed.

Adapt the scope: extend controls for regulated, safety-critical, OT/IoT or geographically constrained environments. This guide cannot replace a specialist assessment of those requirements.

Official Learning References

The topic groupings, questions and illustrative scenarios are editorial guidance written for Atlas. Official sources support the underlying concepts; no third-party article text or diagrams are reproduced.

NIST Cybersecurity Framework

Vendor-neutral outcomes for managing cybersecurity risk.

Open official resource ↗
CISA performance goals

Prioritised practices to inform a context-specific baseline.

Open official resource ↗
Microsoft: revoke user access

Explains account, token, application-session and device considerations.

Open official resource ↗
CISA: known exploited vulnerabilities

Evidence of exploitation to support remediation priorities; not a complete vulnerability inventory.

Open official resource ↗
Microsoft Cybersecurity Reference Architectures

Microsoft-specific architecture guidance using Zero Trust principles.

Open official resource ↗
Microsoft Security Adoption Framework (SAF)

Structured guidance connecting business priorities, security responsibilities and technology controls to plan and prioritise Zero Trust adoption across hybrid and multicloud environments.

Open official resource ↗
AWS: AgentCore security practices

Runtime isolation, identity, permissions, secrets and network boundaries.

Open official resource ↗
AWS: agent prompt-injection security

Layered safeguards for agent inputs and actions.

Open official resource ↗
CISA: StopRansomware Guide

Preparation, protection, response and recovery learning.

Open official resource ↗
Microsoft Defender XDR

Coordinated threat detection and response across supported security components.

Open official resource ↗
Microsoft Defender for Cloud

Cloud posture and workload protection capabilities.

Open official resource ↗
Amazon GuardDuty

Threat detection using supported AWS signals.

Open official resource ↗
Amazon Inspector

Vulnerability assessment for supported AWS workloads.

Open official resource ↗
AWS Security Hub CSPM

Cloud posture checks and consolidated security findings.

Open official resource ↗
Google Security Command Center

Cloud posture and threat visibility; validate tier and workload coverage.

Open official resource ↗
Microsoft Sentinel

SIEM capabilities for collecting signals, investigating and coordinating response.

Open official resource ↗
Microsoft Purview

Data protection and governance; check individual capability coverage and release status.

Open official resource ↗
Microsoft: Insider Risk Management

Potential intentional and inadvertent insider risks; findings need contextual review.

Open official resource ↗
Microsoft: insider risk privacy

Privacy safeguards, pseudonymisation and controlled investigation access.

Open official resource ↗
Microsoft: discover and manage shadow IT

Discover, assess and govern unsanctioned cloud application use.

Open official resource ↗
Microsoft: sensitivity labels

Classification and configurable protection across supported content and services.

Open official resource ↗
Microsoft: sensitive information types

Pattern-based identification; classification detection is distinct from a protection policy.

Open official resource ↗
NIST: secure software development

Secure development and acquisition practices; check the publication status of any revision.

Open official resource ↗
NIST: Zero Trust Architecture

Resource-focused access decisions; network location alone does not establish trust.

Open official resource ↗
Microsoft CAF: enforcing cloud governance policies

Guidance on translating policy into operational controls, with clear ownership, proportionate automation and ongoing review. Azure-specific examples should be adapted to the organisation’s environment.

Open official resource ↗
C2PA: content provenance

Provenance helps assess creation and editing history; it does not prove depicted claims are true. Missing credentials do not prove media is fake.

Open official resource ↗
Microsoft: agent accounts (Frontier preview)

Dedicated accounts, separately managed Agent IDs, tool authentication and preview scope.

Open official resource ↗
Microsoft: shared-memory FAQ (Frontier preview)

This preview has no user-to-user memory privacy boundary. Consult current retention, deletion and opt-out limitations.

Open official resource ↗
Microsoft: Entra Agent IDs for Copilot Studio

Agent identity governance and lifecycle; distinct from the dedicated agent-account feature.

Open official resource ↗
Microsoft Entra security defaults

Preconfigured identity protections and guidance on moving to Conditional Access.

Open official resource ↗
Microsoft Entra Conditional Access

Context-aware access policies; validate licensing and scope.

Open official resource ↗
CISA: Secure by Design

Security throughout product design and development, including secure defaults.

Open official resource ↗
Microsoft: business continuity and disaster recovery

Explains recovery time and recovery point objectives.

Open official resource ↗
Microsoft: business continuity planning

Business impact analysis, dependencies, recovery priorities and continuity exercises.

Open official resource ↗
NIST: cybersecurity event recovery

Recovery planning, playbooks, testing and improvement after cyber incidents.

Open official resource ↗
Microsoft Secure Score

Measures progress against supported Microsoft security recommendations; not a guarantee of security.

Open official resource ↗

CISA’s catalogue is a useful exploitation signal, not a complete risk assessment. Any binding federal remediation deadlines have a defined scope and are not universal deadlines. Frameworks and vendor resources have different roles.