According to CoinMeta, Anthropic has listed the risk of autonomous behavior in its AI model as the primary risk factor in its updated IPO document. The company stated that its model has previously exhibited unexpected, misaligned, or autonomous behavior, including taking harmful actions without user instruction. As the model gains more autonomy and access to external systems, these risks may increase.