Noma Labs Identifies Vulnerability in GitHub's AI Workflows Allowing Data Leakage

Jul 8, 2026

AI Summary

Noma Labs has uncovered a significant vulnerability in GitHub's Agentic Workflows, enabling attackers to extract data from private repositories by exploiting prompt injection techniques. This flaw allows unauthorized users to manipulate the AI agent into revealing sensitive information through crafted GitHub Issues in public repositories.

Noma Labs discovered a prompt injection vulnerability in GitHub's Agentic Workflows, named GitLost. This vulnerability allows unauthenticated attackers to access private repository data by posting crafted issues in public repositories of the same organization.
GitHub Agentic Workflows combine GitHub Actions with an AI agent, allowing teams to automate tasks using natural language in Markdown files. The AI agent can read issues and access other repositories within the organization.
The vulnerability arises from the AI agent treating untrusted content as reliable instructions, leading to potential data leaks. Attackers can exploit this without needing coding skills or credentials, simply by creating a public issue.
Noma Labs demonstrated the vulnerability by crafting a seemingly innocent GitHub issue that triggered the AI agent to fetch and publicly comment on the contents of private repository files.
GitHub had implemented security measures to prevent such leaks, but these failed in this instance due to the AI's misinterpretation of the input.
The findings were responsibly disclosed to GitHub, highlighting the need for improved security measures in agentic AI systems to prevent similar vulnerabilities.

githubai ethicsdata privacysecurityai risks

Noma Labs Identifies Vulnerability in GitHub's AI Workflows Allowing Data Leakage

Related Stories

Teen Social Media Bans Overlook AI Chatbots and Their Impact on Youth

Anthropic appoints Ben Bernanke to its Long-Term Benefit Trust

Ben Bernanke appointed to oversight trust at AI company Anthropic

Meta introduces camera safeguard for AI glasses amid privacy concerns and ongoing scrutiny