Skip to content
NewA calmer, faster ZeroHour

Device-code phishing can abuse GitHub Copilot CLI, Microsoft finds

Original title: The risks of device code authentication for AI coding agents

Microsoft Security Blogby Uri Oren ()11 min read
AI summary · x-ai/grok-4.7Summaries by Vani AI

A victim must approve an attacker-started GitHub device code; Microsoft says the four follow-on abuses were not seen in the wild.

Microsoft describes a lab simulation in which a test account approved a GitHub device-code sign-in started by an attacker-controlled client using GitHub Copilot CLI. Depending on permissions and features, that authorization could steer a remotely controllable session, add misleading agent memory, read cloud-synced sessions, or reach private repositories. Microsoft says these abuse designed features and were not observed in the wild, and points to Defender detections linking phishing to GitHub authentication.

Unconfirmed: Microsoft says the four outcomes are lab proof-of-concept results, not attacks observed in the wild.

Key facts
  • Any public OAuth device-flow client can be abused, and the attacker chooses the scopes.
  • Remote steering needs user-enabled remote control and an allowing organization policy.
  • The verification page is the real GitHub site, so the domain does not identify the initiator.
  • Named earlier activity includes the EvilTokens campaign and Storm-2372.
  • Microsoft says it disrupted EvilTokens with partners in September 2026.
Security assessment · AI, from the article
What to do
  1. 1.Correlate device-code phishing with GitHub sign-ins in Microsoft Defender.
  2. 2.Restrict Copilot CLI remote control to sessions organization policy allows.
Explain this AI news with Vani AIWhat was announced, what is new and why it matters, in a new chat

From the original article

11 min read at microsoft.com

Lab research shows how device code phishing, a technique that applies to any device-flow client, can extend to AI coding agents, using GitHub Copilot CLI as an illustrative example. The post The risks of device code authentication for AI coding agents appeared first on Microsoft Security Blog.

Microsoft Security Blog

By Uri Oren · microsoft.com

Read the full article at Microsoft Security Blog

A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.