TL;DR: AI has moved into the core of Microsoft’s Security stack. Treat agents, identity and data governance as first-class parts of your security architecture in 2025, consolidate where it makes sense and use automation to close the capacity gap.
By Ben Harding, Microsoft Alliance Director at CyberOne
I’m writing this from San Francisco at Microsoft Ignite 2025. The theme is crystal clear. AI is moving faster than most organisations can retool and Microsoft is rebuilding its security platform to match. Identity, data and cloud controls are being reimagined around AI and autonomous agents. If you plan to scale AI, security needs to be in the room from day one.
This is the biggest shift I’ve seen in Microsoft Security for years. Below is my practical take for organisations - what to pay attention to and why it matters for your 2026 plan.
AI is no longer a bolt-on to Microsoft 365. It now runs through the stack. This year, Microsoft focused on securing AI models, grounding data and importantly, autonomous agents.
Security Copilot agents are arriving across Microsoft Defender XDR, Microsoft Sentinel, Microsoft Entra, Microsoft Intune and Microsoft Purview. These agents handle triage, investigation, posture improvement and configuration analysis. In short, they give stretched IT and security teams real help without extra complexity or niche skills.
Microsoft Defender for Cloud adds AI-aware posture management and runtime protection for Azure AI Foundry services. Microsoft Purview adds stronger controls to discover sensitive data used by models, block unsafe prompts, and prevent sensitive content from being used for grounding.
Identity is being extended to non-human actors. Microsoft Entra Agent ID brings conditional access, lifecycle management and full audit to AI agents, tackling agent sprawl before it starts. This will be essential as organisations scale automation and agent-driven workflows.
Microsoft is also releasing unified AI security dashboards that merge signals from Defender, Purview and Entra into a single view of AI assets, risks and usage.
The direction of travel is clear: you cannot adopt AI safely without treating it as part of your core security architecture.
Microsoft 365 Copilot Business is a new offering for organisations with fewer than 300 users. Priced at $21 per user monthly, it gives SMBs access to AI-driven productivity without added complexity. It automates everyday tasks such as email summarisation, document drafting, data analysis and meeting notes.
Microsoft Teams can also add agents into Copilot Business to handle repeatable workflows, freeing staff to focus on customers and strategic work. General availability lands in December.
Entra Agent ID gives every AI agent a unique enterprise identity. Organisations can now discover, govern and secure their entire agent fleet, including shadow or unsanctioned agents.
Baseline Security Mode is now generally available and provides a guided way to harden the Microsoft 365 environment. In a few clicks, IT and security teams can identify configuration gaps, simulate changes and roll out recommended security settings across Office, SharePoint, Microsoft Teams, Entra and more.
For many organisations, this is a direct route to reducing risk quickly without disrupting operations.
Microsoft Sentinel has moved beyond being “just another SIEM”. Microsoft clearly positions it as the central telemetry and AI fabric for modern SecOps.
The platform now includes a deeper data lake, graph insights and improved AI-driven analytics. Threat hunting becomes more visual, more intuitive and less dependent on deep KQL. With 350+ connectors, including Microsoft 365 Defender, Entra ID, Azure services, AWS CloudTrail, Okta, CrowdStrike, Palo Alto and Cisco firewalls, organisations can consolidate logging into Sentinel and retire duplicate ingestion pipelines.
Hands-on sessions at Ignite focused on AI-assisted hunting, natural-language investigation and agent-driven analysis. For security teams, this means faster investigation and shorter response times without adding headcount.
If you are still running multiple SIEM or log tools, this is a good moment to revisit the strategy.
Microsoft Defender for Cloud continues to mature into a strong multicloud security and posture platform.
Microsoft Cloud Security Benchmark v2 adds expanded controls, better alignment with Azure Policy and fresh guidance for securing AI workloads. This gives you a more structured and measurable way to improve cloud posture.
Ignite Labs showed how Defender for Cloud connects misconfiguration detection, attack path analysis and integrated response across Azure and AWS. It is clear Microsoft intends Defender for Cloud to be the anchor for cloud security and governance.
For growing cloud footprints or hybrid estates, this makes it easier to standardise posture management.
Identity has long been the backbone of Microsoft Security. What changes this year is the scope.
Agent ID provides identities for autonomous agents so every non-human actor has the same governance, control and audit trail as a human user. This matters as you adopt low-code agents, task bots and M365 Copilot automations.
Entra Internet Access for AI adds network-level protection for AI usage, allowing you to control and monitor how AI services are accessed. Combined with strengthened identity protection, risk-based access and improved app governance, Entra offers a more unified approach to secure access in the AI era.
As autonomous workflows grow, a reliable identity for humans and agents becomes non-negotiable.
Preview features designed to strengthen protection and reduce friction:
Data protection is where AI security becomes very real. Microsoft Purview now offers stronger Data Security Posture Management, guided investigations and extended DLP for AI interactions.
The ability to block unsafe AI prompts or prevent sensitive information from being used as grounding data is a major step for regulated industries. Purview also extends protection across browsers and networks, giving tighter control over how sensitive data moves between SaaS platforms and AI tools.
For finance, healthcare, life sciences and professional services, these data-centric controls will be central to any AI adoption strategy.
Purview DLP for Copilot prevents sensitive information from being used in prompts or grounding data.
This preview feature identifies and fixes overshared links at scale, reducing unnecessary data exposure and improving compliance posture.
AI Observability provides visibility into agent activity across Microsoft 365, helping teams understand how agents interact with data and proactively manage risk.
I see 3 near-term priorities:
Copilot agents and natural language workflows will reduce pressure on operations. This is where organisations can realise tangible gains in 2025.
Ignite 2025 confirms that AI and security are now the same conversation. Microsoft is delivering a unified platform that helps organisations adopt AI safely, retain control over their data, and strengthen identity governance as autonomous agents scale.
For organisations, this is the time to modernise security without adding complexity and to make AI work for your people, not against your operational capacity.
At CyberOne, we exist to help leaders do exactly that - combining Microsoft Security with world-class Consulting, Professional and Managed Services so you can move from risk to resilience with measurable outcomes and guaranteed SLAs.
Join Our Post-Ignite Security Briefing
We will break down what matters, what to prioritise and how to put these capabilities into action in 2025.
Register Now To Secure Your Place.