September 2022

September 2022 is a brief but pivotal month, capturing the earliest documented discussions of existential AI risk alongside the first notable research on prompt injection attacks against GPT-3. Meta's text-to-video generation and the DreamFusion 3D model from Google signal rapid multimodal expansion beyond static image generation, suggesting that the generative AI frontier was moving far faster than safety discourse could track. The small number of entries belies their significance — this is the month the vocabulary of AI security and AI existential risk begins to crystallize in public conversation.