Tools AI Risk Radar ai-incident-0172
Incident record
Copilot CLI test leaks secrets through encrypted web content
- Severity
- High
- Status
- Proof-of-concept
- Type
- Prompt Injection
- Target
- GitHub Copilot CLI in autopilot mode
- Actor
- researcher
What happened
Adversa demonstrated injected instructions concealed in encrypted web content. When a user asked the agent to read the attacker-controlled page, the tested model could decode the content and send local secrets away.
The researcher reports success in 50% of runs with mai-code-1.1-flash; two tested GPT-5.6 variants refused. GitHub validated the behaviour but declined to classify it as a vulnerability. This record covers the controlled demonstration.
Sources
- Primary disclosureadversa.ai/blog/cryptographic-context-injection-github-copil…
One record from the AI Risk Radar, maintained by the Security Desk.Data: CSV · JSON ·RSS · CC BY 4.0 with attribution to YFarmX.Tracker updated · 13 October 2026