GitHub puts Copilot in US/EU data walls and clears FedRAMP

GitHub just moved one of the biggest enterprise blockers around coding copilots from procurement conversation to shipping product. In a April 13 changelog post, the company says GitHub Copilot now supports data residency in both the US and EU, keeping inference processing and associated data inside the selected geography. For US government customers, GitHub also says the model hosts and supporting infrastructure now meet FedRAMP Moderate authorization standards.

This is more consequential than a narrow storage-location update because GitHub is applying it across the full generally available Copilot surface, not just chat. The post says agent mode, inline suggestions, chat, Copilot cloud agent, code review, pull request summaries, and Copilot CLI all run through data-resident, compliance-certified endpoints in the chosen region. That turns residency from a partial checkbox into a policy enterprises can actually enforce across day-to-day developer workflows.

Model choice matters here too. GitHub says the launch includes OpenAI and Anthropic models such as GPT-5.4, Claude Sonnet 4.6, and Claude Opus 4.6, with documentation covering the exact model-by-region matrix. One notable exclusion is Gemini: GitHub says Google Cloud still does not provide data-resident inference endpoints for those models, so Gemini is out until that changes. That leaves enterprises with a real menu, but not a complete one, which matters for teams that have standardized around specific model behavior.

There is also a cost and governance story hiding behind the compliance banner. Data-resident and FedRAMP requests come with a 10% increase in the model multiplier, meaning a request that normally costs 1 premium request is billed as 1.1 under these controls. The policies are off by default, so enterprise and organization admins have to actively enable the restrictions in Copilot settings. GitHub is effectively saying compliance is available, but customers need to accept both the narrower model set and the pricing trade-off.

Why this matters right now is simple: AI coding tools are moving from experimental developer perk to something enterprises want inside regulated engineering environments. Data residency and FedRAMP are rarely headline-grabbing features, but they decide whether a product reaches public-sector teams, financial firms, healthcare groups, and multinationals with stricter geography rules. GitHub is also signaling where the next expansion goes, with Japan and Australia listed as later 2026 targets. The bigger takeaway is that Copilot competition is no longer just about model quality; it is also about who can package frontier models inside enterprise-grade policy controls.