GitHub push protection for AI-detected secrets

GitHub push protection now uses a context-aware classifier, built with Microsoft, to block secrets with no predefined pattern.

GitHub · Security and safety tooling · security

Published in Feed

Overview

GitHub announced on October 28, 2026 that push protection can block secrets that match no predefined pattern, using a classifier built with Microsoft Applied Sciences that reads how a candidate secret is used in the surrounding code. It runs next to the pattern-based detectors and covers pushes, commits made in the GitHub UI and file uploads; in early results it more than doubled the number of pushes where push protection caught a secret, with precision GitHub's users report as close to partner-defined detectors. It is in public preview for organizations with GitHub Secret Protection or GitHub Advanced Security on github.com, off by default, and uses AI credits once enabled. Agents must get authorization before bypassing a block.

This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.

Key features

  • Context-aware classifier detects secrets without a known format
  • Runs alongside pattern-based detectors on pushes, UI commits and file uploads
  • Early results: more than doubled pushes where a secret was caught (GitHub-reported)
  • Enabled via repository settings or an org/enterprise security configuration
  • Agents need authorization to bypass or work around a block

Use cases

  • Catching custom tokens and internal credentials that have no public pattern
  • Stopping leaked secrets from coding-agent commits before they reach history
  • Tightening secret protection across an enterprise via security configurations

Pricing

Public preview for GitHub Secret Protection or Advanced Security customers; AI detection in push protection uses AI credits once enabled. Post-push AI-detected alerts stay included at no extra charge.

Pricing reflects the source record and may change. Confirm current terms with the provider.

Dates and sources

Released
28 Oct 2026
Record updated
2 Nov 2026
Record last checked
2 Nov 2026

Explore the Feed

Other entries filed under security and safety tooling; inclusion does not imply equivalent features.

  • DefenseClaw Edge Connector

    Cisco's open-source DefenseClaw gains an Edge Connector that allows or blocks AI agent tool calls directly on IoT and robot hardware.

  • Lakera Guard 2.0

    Lakera's real-time guardrail API for LLM apps, updated with multi-agent protection and a new benchmark for detecting indirect prompt injection in tool outputs.

  • Fleming-1

    Sierra's model that listens to a live phone call and scores whether the caller is an AI agent rather than a human.

Browse the Feed