← Incident Database
Indirect Prompt InjectionCritical

CamoLeak: private source-code exfiltration from GitHub Copilot Chat

October 2025 · GitHub Copilot Chat
What happened
Hidden prompts embedded in pull-request descriptions could steer Copilot Chat to leak source code and secrets from private repos. Exfiltration abused GitHub's own Camo image proxy, encoding stolen data as a sequence of pre-generated image requests (CVE-2025-59145, CVSS 9.6).
Root cause
Copilot acted on untrusted PR content as instructions, combined with a CSP bypass that allowed data egress through the trusted Camo image path.
Fix / outcome
GitHub disabled image rendering in Copilot Chat and blocked the Camo exfiltration path in mid-August 2025.
Sources
Learn this attack class
This incident is an example of Indirect Prompt Injection. Read the guide, then try it hands-on in the Academy.
Read the guide →Try the challenge
← Back to the Incident Database