GitHub AI Secret Detection: Included Alerts and Planned Credit Charges

Existing password alerts get the model upgrade without an added charge

The purpose-built model identifies likely leaked credentials using surrounding code, including passwords without a recognizable token format. It is a fine-tuned detection classifier, rather than a model for generating code or prose. Broader detection does not guarantee that it will find every secret.

GitHub says customers using AI-detected Password alerts have automatically moved to the new model. Existing AI-detected secret-alert scans remain included with GitHub Secret Protection (GHSP) and GitHub Advanced Security (GHAS) at no additional charge. That is a separate billing path from the new optional checks that will consume Credits.

The new checks have preview and billing plans, not one universal start date

The October 7 announcement covers three stages: existing alerts, push-time protection and Copilot security reviews. At that announcement, AI push protection was in private preview. Added classifier checks for /security-review in Copilot CLI and the Copilot app were described as coming soon in private preview.

GitHub planned to introduce AI Credits usage for the new opt-in checks in the following weeks and identifies joining the public preview and enabling the feature as billing triggers. It gives no exact effective date, per-check price or Credits multiplier. Check the relevant preview notice for actual access and billing timing rather than treating the October 7 plan as proof of a later completed rollout.

Push protection: identify the repository owner and the EMU exception

AI push protection checks for unstructured credentials at push time, offering a chance to remove them before they enter repository history. On github.com, GitHub Team or GitHub Enterprise Cloud customers need paid GHSP/GHAS coverage. Public, private and internal repositories can qualify, with an administrator enabling the feature under organization or enterprise policies.

Planned usage normally belongs to the repository-owning organization and appears under the Secret Protection AI Credits SKU in usage insights, without consuming a particular user's allocated Credits. User-namespace repositories for enterprise-managed users (EMUs) are an explicit exception: usage belongs to the pusher and uses that person's allocated Credits. A check can consume Credits even when it does not block a push.

Copilot security review: new checks add usage but are not enabled by the command

In a supported Copilot CLI or Copilot app session, /security-review examines active changes and returns prioritized security findings with remediation suggestions. The existing review is read-only and follows existing Copilot policies and billing. Acting on a finding is a subsequent operation that needs the appropriate authorization.

The planned secret-classifier checks do not require a GHSP/GHAS license, but will consume AI Credits in addition to the review's existing usage. The active Copilot plan's billing account receives that usage, reported under GHSP in usage insights. Do not confuse this billing account with the repository-owning organization used for push protection.

The new checks are off by default. Running /security-review alone does not enable them: users must opt in and enable them where their plan and policies allow. Individual opt-in does not override organization controls. GitHub also says agents should not enable credit-consuming features or change policies or budgets without explicit authorization.

Separate plan and platform requirements: Copilot does not replace GHSP

Individual Copilot Pro, Pro+, Max, Free and Student plans can be eligible for the new security-review checks, subject to access controls and credit consumption. Business and Enterprise access on supported platforms also depends on invitations and administrator policies. Eligibility for Copilot review checks does not supply the GHSP/GHAS coverage required for AI push protection.

For GitHub Enterprise Cloud with data residency on ghe.com, the announcement names Copilot Business and Enterprise as supported plans for the security-review command. AI push protection there remains planned with paid GHSP/GHAS coverage. GitHub hosting plans, GHSP licenses and Copilot subscriptions are separate layers; sharing the Enterprise name does not make them interchangeable.

GitHub Enterprise Server (GHES) 3.23 is planned to bring AI-detected alerts in public preview, included with an existing GHSP/GHAS purchase. AI push protection and the Copilot security-review command are not part of that Server release. The announcement also gives no comprehensive minimum client-version list, country availability table or per-account invitation schedule.

Budget setup: an alert and a usage stop are different controls

First, identify the organization or enterprise administrator authorized to manage the spending and the applicable feature policies. Administrators can disable the new capabilities and set AI Credits budgets. Included credits and additional paid usage still follow the account's billing policies and limits.

Second, follow the announcement's path: open Billing and licensing, then Budgets and alerts. Choose SKU-level budget, select Advanced Security as the product and Secret Protection AI Credits as the SKU. An all AI Credits budget can instead cover multiple credit-consuming SKUs.

Third, distinguish notification from enforcement. Budget alerts alone do not stop usage. For a spending cap, configure Stop usage when budget limit is reached where that option is available, rather than treating an alert as proof that billing has stopped.

Fourth, teams already in the AI push-protection private preview should watch the billing-change notice. GitHub warns that continuing after the change takes effect will consume AI Credits; disable the feature beforehand if that usage is unwanted. Existing AI alert scans remain included without an additional charge, so avoiding the new charges does not mean every alert capability must be disabled.

A hypothetical team can have two different billing owners

Suppose an organization owns a team's repository and a developer has an active Copilot plan. First list whether the intended capability is an existing alert, AI push protection or the new security-review classifier. Then check GHSP/GHAS coverage, Copilot eligibility and preview invitations separately. This hypothetical decision process does not mean the previews are already available to that team.

If the team later qualifies and opts into the new checks, push protection normally belongs to the repository-owning organization, while security review follows the active Copilot billing account. Apply the separate exception for an EMU user-namespace repository. Establish the payer, budget-stop settings and administrator authorization before enabling or continuing a preview; no blocked push, no finding or a budget alert alone does not prove there was no usage.