LLMMedium
GPT-3-Based Twitter Bot Hijacked Using Prompt Injection Attacks
- Occurred
- Sep 15, 2022
- Published
- Sep 19, 2022
- Source
- AI Incident Database
Summary
Remoteli.io's GPT-3-based Twitter bot was shown being hijacked by Twitter users who redirected it to repeat or generate any phrases.
Developers
- OpenAI
- Stephan de Vries
Deployers
- Stephan de Vries
Harmed parties
- Stephan de Vries
Classification
Category confidence: 95%
Severity confidence: 90%
Rationale: GPT-3 bot hijacked via prompt injection, representing a language model security failure.
Classifier: Qwen/Qwen3.5-35B-A3B-FP8