15 October 2026
When adopting Microsoft Copilot and AI-powered agents, organisations may focus on productivity gains while underestimating the associated security and governance challenges. Unlike traditional software, AI creates new ways for users to discover information, interact with business systems and automate activities across the organisation.
Key security considerations include the following:
Data access and oversharing - Copilot operates using the permissions that users already possess and can surface content from emails, documents, chats and SharePoint sites that a user is authorised to access. Organisations should assess and remediate excessive permissions before deployment. Copilot can make existing access significantly more visible, exposing historic oversharing and poor information management practices.
Identity and access management - As AI becomes embedded in business processes, protecting identities becomes even more critical. Organisations should implement strong authentication, Conditional Access policies, least-privilege access principles, and a Zero Trust approach to ensure that only authorised users can access sensitive information through Copilot capabilities.
Data protection and leakage prevention - Organisations should use Microsoft Purview sensitivity labels, encryption and Data Loss Prevention (DLP) controls to protect sensitive information and ensure that established policies continue to apply when AI systems process or reference content.
Agent governance and lifecycle management - Without clear controls over who can create, publish and manage agents, organisations risk losing visibility and oversight of what agents exist, the information they can access and the activities they perform. A defined lifecycle should cover approval, ownership, access review, monitoring and retirement so that agents remain accountable and appropriately controlled.
Third-party integrations and connected systems - Agents are likely to interact with HR, CRM, finance and other business applications. To reduce the risk of unintended data exposure, organisations should understand what data is available through these connections and ensure that appropriate controls are in place.
Monitoring, audit and compliance processes - AI platforms are evolving rapidly, with new features, settings and capabilities introduced regularly. Maintaining security over time requires ongoing monitoring of Copilot usage, agent activity, data exposure risks, and compliance requirements. Audit trails and governance reporting help organisations demonstrate accountability and investigate potential incidents. Organisations that establish strong controls around data, identities, agents and compliance from the outset will be best positioned to realise the benefits of AI while maintaining trust, security and operational resilience.
We recommend addressing these considerations before deployment, and review them regularly as Copilot capabilities, connected systems, and organisational requirements evolve.
The Kaon Security Microsoft 365 security review incorporates a thorough assessment of a client’s Copilot configuration. Selecting appropriate Microsoft 365 and Copilot security settings can be challenging because of the range and depth of available options. The service helps organisations improve their IT security posture by optimising their Microsoft 365 and Copilot security configuration.
Learn more about our Microsoft 365 Security Review service or contact us today to talk to one of our consultants.
The Kaon Security team are looking forward to exhibiting at this year's LGIT Conference in Coffs Harbour and catching up with councils and local government professionals from across Australia.
If you're attending, swing by our stand (no. 25) and say hello. Whether you have a specific cyber security challenge you're working through or simply want to chat about what's happening across the sector, we'd love to connect.