Microsoft Copilot
Microsoft's Copilot brand spans several materially different products — GitHub Copilot (coding), Microsoft 365 Copilot (embedded in Word, Excel, Teams, and Outlook), and Copilot Studio (building custom agents) — each with its own data-handling and governance profile. Treating "Copilot" as one thing is itself a governance risk.
Last reviewed: 2026-08-24
What it is
"Copilot" is a brand covering distinct products, not a single tool: GitHub Copilot is a code-completion and, in its agent modes, agentic coding assistant; Microsoft 365 Copilot is embedded directly into Word, Excel, PowerPoint, Teams, and Outlook, and grounds its answers in an organisation's own content via Microsoft Graph; Copilot Studio is a platform for building custom conversational agents on top of organisational data and workflows. A governance decision about one does not cover the others.
Where it's adopted
GitHub Copilot is widely adopted by engineering teams for code completion and, increasingly, for agentic coding tasks. Microsoft 365 Copilot is adopted by knowledge workers directly inside existing Office applications, which is precisely what makes it different to assess: it surfaces organisational content the user already has access to, rather than operating on data supplied to it directly. Copilot Studio is adopted by teams building bespoke internal agents rather than using an off-the-shelf assistant.
Governance, compliance, security, and engineering considerations.
Because "Copilot" spans distinct products, sign-off needs to name which product is in scope. A blanket "Copilot approved" decision is imprecise and risks silently approving products with materially different risk profiles under one label.
Microsoft 365 Copilot's grounding in organisational data via Microsoft Graph means existing data-permissions hygiene directly determines what Copilot can surface — a longstanding permissions misconfiguration becomes an AI-surfaced disclosure risk, not merely a dormant one. Data-processing terms follow Microsoft's enterprise agreements and differ by product and licensing tier.
For GitHub Copilot in agentic modes, the same excessive-agency and prompt-injection risk categories apply as any agentic coding tool. For Microsoft 365 Copilot, the notable risk is oversharing: Copilot surfacing content a user technically has permission to see but practically shouldn't — an amplification of existing permission hygiene problems, not a new attack vector.
For GitHub Copilot's agent modes, the same lifecycle control points (human-in-the-loop review, version control, monitoring) as any agentic coding tool apply. For Microsoft 365 Copilot, the operative engineering discipline is the tenant's data and permissions architecture, not code delivery — an access-control audit is the real prerequisite, not a model configuration setting.
Where this connects to the Risk & Control Library.
Informational cross-references, not formal platform tags — none of these entries are specific to Microsoft Copilot alone; they're the generic risks and controls most relevant to this category of tool.
- Sensitive information disclosure — Microsoft 365 Copilot's headline risk — it surfaces what existing permissions already allow, so an access-control gap becomes an AI-surfaced one.
- Excessive agency — Applies to GitHub Copilot's agentic coding modes specifically, not to Copilot as a brand.
- Inadequate AI vendor due diligence — Due diligence needs to be scoped per Copilot product, given how differently they handle organisational data.
- Human-in-the-loop review — For GitHub Copilot's agent modes: approval checkpoints for higher-consequence code changes.
Deployment-governance patterns.
What good practice looks like when adopting this tool.
- Name the specific Copilot product in any governance decision — "GitHub Copilot," "Microsoft 365 Copilot," or "Copilot Studio" — never "Copilot" alone.
- Audit data-permissions hygiene in Microsoft 365 before enabling M365 Copilot broadly — it surfaces what access controls already allow, so it inherits every existing gap rather than introducing a new one.
- Treat GitHub Copilot agent-mode changes with the same version-control and review discipline as any agentic coding tool.
- Confirm which licensing tier and enterprise agreement terms apply to the specific Copilot product in use, rather than assuming one Microsoft-wide data-processing answer.