According to CoinMeta, in the IPO prospectus updated on November 2nd, AI was warned that it might soon cross a critical threshold related to "recursive self-improvement" (RSI), which could lead to a significant acceleration in the research and development (R&D) speed of AI. In the past, there have been instances where significant R&D progress was not detected in time. Currently, Claude is involved in writing code, running experiments, and processing training data to help develop a stronger next-generation model, and this new model, in turn, can accelerate subsequent R&D, creating a cycle of continuous improvement. According to Anthropic's security policy, there are two scenarios in which this critical threshold is reached: 1. AI is able to complete the work of the entire AI research team at a similar cost; or 2. It could double the rate of technological progress, compressing what would normally take two years into one year. As AI delves deeper into its own R&D, the model's capabilities may grow faster than the updates to security assessments, alignment research, and protective measures. Anthropic has not yet identified any existing model that has reached this level. This warning is mainly based on internal estimates and predictions for future models, and they are currently considering modifying the security policy to ensure that critical R&D processes remain under human supervision and control.