Every week I hear the same refrain: “We need Copilot for Microsoft 365 to hit our AI goals.” Here’s the uncomfortable truth for defense missions and the DIB: IL4/IL5 workloads can’t deploy Copilot for Microsoft 365 today based on what’s publicly documented. PubSecAI’s review lays it out plainly: GCC is FedRAMP Moderate and can plan deployments consistent with OMB M-24-10; GCC High and DoD are FedRAMP High aligned to DoD SRG IL4/IL5, and there’s no public listing or documentation showing Copilot availability there, nor a separate FedRAMP authorization for “Copilot” on the Marketplace. Microsoft says Copilot inherits M365 controls and doesn’t train on your tenant data. That’s fine—but inheritance isn’t the gating factor. The IL and ATO are.
In acquisition terms: stop writing requirements that presuppose a capability you can’t legally or technically operate at your impact level. I’m seeing PWS drafts that say “Contractor shall enable Copilot for M365 across the IL5 tenant” like it’s a toggle. It isn’t. And buying licenses doesn’t conjure an authorization. If a vendor tells you “it’s coming next quarter,” treat that as marketing until it’s on the FedRAMP Marketplace and in your ATO package. Hope is not a control.
For programs on GCC (FedRAMP Moderate), you do have room to move—but do it like adults. Adding Copilot changes data access patterns, audit surfaces, and user behavior. That’s a significant change for many SSPs even if it inherits M365 controls. Align with OMB M-24-10: inventory use cases, document risks, apply DLP and sensitivity labels, and scope pilots to non-CUI data sets until you can prove guardrails actually bite in practice.
For IL4/IL5 missions and DIB environments, the risk isn’t abstract. CUI in a Moderate enclave is a compliance failure. I’ve already seen contractors try to “pilot” Copilot by spinning up GCC side tenants and syncing “sample” data. That’s how spills happen. If your contract or CMMC posture requires IL4/IL5 handling, keep CUI and mission data out of features that aren’t available at your IL—full stop.
The acquisition reality:
- Licenses are easy to buy and hard to use. Put options and contingencies in your ordering docs. Do not obligate funds for unavailable features without clear off-ramps.
- Don’t let “use Copilot” creep into award fee criteria or delivery milestones for IL4/IL5. Measure outcomes (productivity, defect rates) without prescribing a tool you can’t deploy.
- Treat Copilot as a configuration-specific addition to your ATO. If you’re in GCC, coordinate with your AO and update artifacts; if you’re IL4/IL5, ensure it’s disabled in policy and technically blocked.
- For primes and subs, write contract language that prohibits routing CUI/export-controlled data through any generative AI assistant not authorized for the required IL, and require the contractor to attest to tenant controls that enforce that prohibition.
What to do now
- If you’re GCC: plan bounded pilots under M-24-10. Update SSPs, enforce DLP/sensitivity labels, and monitor audit logs specific to Copilot interactions.
- If you’re IL4/IL5: document non-availability, disable Copilot features in admin policy, and remove tool-specific language from PWS/SOW. Use option CLINs tied to documented availability (Marketplace listing and vendor documentation for GCC High/DoD).
- For DIB and CMMC: re-affirm that CUI stays in IL4/IL5. Prohibit data syncs or cross-tenant shortcuts to enable Copilot in Moderate environments.
- Vendor management: require written representations on authorization status at your IL and a plan of action when availability is documented. No PowerPoint roadmaps—show me the Marketplace entry and service documentation.
What to watch
- FedRAMP Marketplace for any listing that explicitly covers “Copilot for Microsoft 365.”
- Public Microsoft documentation for GCC High/DoD availability details, not just “inherits controls.”
- SRG and OMB M-24-10 implementation memos that may add AI assistant guardrails.
- Your own ATO change control boards—Copilot is not “just another toggle.”
*Dana Cole is a PubSecAI editorial persona — an AI-generated voice written to represent practitioner perspectives in the defense sector. Views expressed are analytical commentary, not official guidance. *