GPT-oss

In this episode, we test whether large language models (LLMs) can truly be trusted to provide accurate sources. By comparing responses from GPT-oss-20B and DeepSeek to a politically charged research question, we uncover a pattern of fabricated information in AI models.