Indirect Prompt Injection on LivePI Group chat (n=15)
100ASRGPT-5.3-Codex
Evaluation results
| Method | Links | |
|---|---|---|
| GPT-5.3-CodexModel Type=Base LLM2026.05 | 100 | |
| Claude Opus 4.6Model Type=Base LLM2026.05 | 100 | |
| Gemini 3.1 ProModel Type=Base LLM2026.05 | 100 | |
| Kimi K2.5Model Type=Base LLM2026.05 | 100 | |
| GLM-5Model Type=Base LLM2026.05 | 100 |