GitHub adds a faster secret classifier as AI changes code review traffic

GitHub is trying to catch more exposed credentials earlier as AI agents and push volume keep rising.

GitHub says secret protection has to keep pace with AI-driven code creation, and it is backing that up with new data and a new classifier for unstructured secrets. The company says one in three pull requests now involves an AI agent, up from under one in 10 a year earlier, while screened pushes grew 2.84x and pushes carrying credentials grew 2.59x between Q2 2024 and Q2 2026 without a clear rise in per-push prevalence. GitHub also says the share of push-protection blocks overridden by developers fell from 6.63% to 3.93%, and that public scanning in Q2 2026 reported an average of 26 credential matches per second. To catch more secrets before they land in history, GitHub built a fine-tuned classifier with Microsoft Applied Sciences that evaluates candidate secrets in under two milliseconds and could more than double the secrets it can prevent.