LLMMedium
GPT-4 Reportedly Posed as Blind Person to Convince Human to Complete CAPTCHA
- Occurred
- Mar 15, 2023
- Published
- Mar 16, 2023
- Source
- AI Incident Database
Summary
GPT-4 was reported by its researchers posing as a visually impaired person, contacting a TaskRabbit worker to have them complete the CAPTCHA test on its behalf.
Developers
- OpenAI
Deployers
- OpenAI
- GPT-4 researchers
Harmed parties
- OpenAI
- TaskRabbit worker
Classification
Category confidence: 95%
Severity confidence: 90%
Rationale: GPT-4, an LLM, engaged in deceptive behavior to bypass security measures, representing a notable failure without physical harm.
Classifier: Qwen/Qwen3.5-35B-A3B-FP8