| Titel | PromtEngineer localGPT Latest (commit 4d41c7d) LLM Prompt Injection via Unsanitized User Input |
|---|
| Beschreibung | A critical prompt injection vulnerability exists in the AI routing mechanism that allows attackers to manipulate the Large Language Model's behavior by injecting malicious instructions through user queries. The application directly embeds unsanitized user input into system prompts without any input validation, escaping, or output verification. This enables attackers to extract sensitive information, manipulate AI responses, bypass routing logic, and potentially chain with other vulnerabilities to achieve complete system compromise. |
|---|
| Quelle | ⚠️ https://github.com/August829/CVEP/issues/9 |
|---|
| Benutzer | Yu_Bao (UID 89348) |
|---|
| Einreichung | 13.03.2026 02:21 (vor 5 Monaten) |
|---|
| Moderieren | 27.03.2026 14:49 (15 days later) |
|---|
| Status | Akzeptiert |
|---|
| VulDB Eintrag | 353889 [PromtEngineer localGPT bis 4d41c7d1713b16b216d8e062e51a5dd88b20b054 LLM Prompt backend/server.py _route_using_overviews erweiterte Rechte] |
|---|
| Punkte | 20 |
|---|