GitHub Enterprise Server 3.23 release candidate adds richer Projects filters and on-appliance AI secret detection
GitHub Enterprise Server 3.23 RC adds compound Projects filters and previews on-device AI secret detection, including for air-gapped deployments.

- GHES 3.23 RC lets users build more precise GitHub Projects views with `OR`, explicit `AND`, and parentheses for grouping.
- GitHub says AI secret detection can run fully inside the customer network, including air-gapped environments, with no GPU, hosted LLM, or outbound connection.
- The AI secret detection feature is in public preview, so GitHub is still warning that the behaviour and details may change.
- The announcement is a release-candidate note, not a full general-availability release, so admins should treat the details as subject to final packaging and rollout decisions.
GitHub Enterprise Server 3.23 release candidate is out
GitHub has announced the release candidate for GitHub Enterprise Server 3.23, and the changelog highlights two items that will matter to many enterprise teams: more expressive filtering in GitHub Projects and AI secret detection that runs on the appliance itself. The release candidate landed on 2026-11-20, according to GitHub’s changelog. GitHub announcement
What’s new in GHES 3.23 RC
The clearest Projects change is syntax. GitHub says users can now build more precise project views by combining filters with logical OR, explicit AND, and parentheses for grouping. In the example GitHub gives, (status:Done OR status:Todo) is:issue returns issues in either status. That is a small change on paper, but it gives admins and project owners a more expressive way to slice large boards without flattening conditions into separate views. GitHub announcement
GitHub also says enterprises using GitHub Secret Protection can detect unstructured secrets entirely within their own network with AI secret detection on GHES, including in air-gapped environments. The company says the model is self-contained, fine-tuned, and runs on the appliance CPU without a GPU, hosted LLM, or outbound connection. GitHub labels the feature as public preview and says it is subject to change. GitHub announcement
Why it matters
For developers, the Projects update reduces friction when a single view needs to cover more than one status or issue type. The practical value is in fewer workarounds: instead of splitting conditions across multiple views, teams can use grouped logic to match how they actually triage work. GitHub announcement
For security and platform teams, the secret-detection note is the bigger architectural point. GitHub is explicitly describing an on-premises inference path for customers that cannot or do not want to send content outside their network. The air-gapped support, CPU-only runtime, and lack of outbound connection are the important constraints here. Because the feature is still in public preview, teams should treat it as something to evaluate, not a finished default. GitHub announcement
At a glance
| Area | Confirmed change | What to watch |
|---|---|---|
| GitHub Projects | Filter logic can use OR, explicit AND, and parentheses |
Whether complex views become easier to maintain at scale |
| AI secret detection | Runs entirely within the customer network, including air-gapped setups | Public preview status and any change before GA |
Availability and caveats
GitHub’s post is a release-candidate announcement, so it confirms what is being introduced in GHES 3.23 but does not by itself settle every deployment detail. The changelog does not give pricing, final GA timing, or rollout mechanics in the passages provided here. It does, however, make the feature boundaries clear: the Projects filtering enhancement is available in the 3.23 RC announcement, while AI secret detection is in public preview and still subject to change. GitHub announcement
If you administer GHES, the sensible next step is to review whether your current Projects views depend on awkward filter combinations and whether your Secret Protection rollout could benefit from an on-appliance detection path for unstructured secrets. Both items are worth checking now, but only one of them — the AI secret detection capability — comes with an explicit preview warning from GitHub. GitHub announcement
GitHub Projects filtering
Combine filters with `OR`, explicit `AND`, and parentheses for grouping.
AI secret detection
Runs on the appliance CPU inside the customer network, including air-gapped environments; public preview.
