GitHub’s changelog of 13 November says the AI Scan for repositories setting in code scanning can now be turned on or off through the REST API, at both organisation and repository level. The feature is in public preview.
The facts
- A new
repository_scanproperty sits alongside the existingpr_scansetting. It acceptsenabledordisabled. - If an organisation disables it, its repositories cannot enable it. If an organisation enables it, repositories inherit that and can opt out individually.
- If the enterprise policy does not allow AI Scan, enable requests are rejected.
- Enabling does not start a scan. People with write access run scans on demand, and each scan uses GitHub Code Security AI credits.
- It requires GitHub Code Security or GitHub Advanced Security for private repositories. GitHub Enterprise Server is not supported.
What it means
The headline is control. Security tooling that uses AI is no longer a switch someone flips repo by repo; it can be managed by script across an estate. The detail that matters commercially is the credit usage. AI-powered checks are metered, so the cost of “scan everything” grows with how often your team presses the button.
This is part of a wider pattern in AI development tooling: features arrive with usage-based billing and admin controls attached, and budgets need an owner.
So what
If AI is writing more of your code, someone needs to be checking it, and that check has a price. Ask whoever builds your software how AI-generated code is scanned, who can trigger scans, and where the credit spend is reported. A clear answer is a good sign about how seriously they treat security.
BuildApps combines AI-assisted delivery with human review. Read about AI-assisted development or start a project.