On 17 November 2026 GitHub updated Copilot code review with several changes. Two matter most to anyone paying for software built with AI help.
Adaptive review effort (public preview). Copilot assesses each pull request’s complexity and risk, then reviews it at Lite, Balanced or Max effort and shows which level it chose. Balanced stays the default. Adaptive never exceeds the maximum effort set for the enterprise, organisation or repository.
Secret detection (public preview). Copilot flags exposed secrets, such as keys and passwords, in review comments before code is merged. GitHub says it matches over 500 known secret types from more than 150 partners, and uses AI models to catch secrets with no recognisable pattern. It is switched on in repository or organisation settings.
Also announced
- Cross-repository context (public preview, public repositories only), so a review can read related public code.
- Review outcomes in Repository insights, showing how Copilot feedback is acted on.
- Copilot code review for Azure DevOps is generally available, billed in GitHub AI credits to the linked Azure subscription.
The preview features are on Copilot Pro, Pro+, Max, Business and Enterprise, and GitHub notes preview features may change.
Why it matters
AI now writes a growing share of code, so the bottleneck is checking it. Leaked credentials are one of the most expensive mistakes a young product can make, and automated checks are becoming a normal part of the pipeline. Effort levels also mean review has a cost dial: more thorough means more spend.
So what
Ask any development partner what automated review and secret scanning run on your code, and who reads what they flag. Our AI-assisted development process pairs automated checks with human review. For bespoke builds, see custom software development.