The video explores a "Semantic Attack" on Google's Gemini AI. This vulnerability exists at the interpretation layer, not the code level. Traditional attacks use keywords. Semantic attacks use how LLMs like Gemini understand context to bypass safety measures.
