← Incident Database
Indirect Prompt InjectionHigh

GitHub Copilot Chat prompt injection to data exfiltration

June 2024 · GitHub Copilot Chat
What happened
Hidden instructions in untrusted source code that Copilot analyzed could fully control its responses and exfiltrate data by rendering an image tag whose URL carried stolen data to a remote server.
Root cause
Copilot Chat followed injected instructions from analyzed code and rendered remote images, allowing data to leak via the image-request URL.
Fix / outcome
GitHub mitigated by disabling image rendering in Copilot Chat. (Distinct from the more severe 2025 CamoLeak issue.)
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